Merge lp:~julian-edwards/maas/1.7-race-for-staticip-bug-1387262 into lp:~maas-committers/maas/trunk

[Julian Edwards]

Backport trunk r3337: Use a lock in the critical section of the code that finds and records a new static IP address in the database. Some users have reported errors in the PG log file complaining about inserting duplicates, so there must be a race.

[Julian Edwards (julian-edwards)]

Approved for a backport by Andres earlier in the bug.

review: Approve

=== modified file 'Makefile'
--- Makefile	2014-10-01 07:22:16 +0000
+++ Makefile	2014-11-06 01:24:10 +0000
@@ -351,8 +351,8 @@
 # has a bug and always considers apt-source tarballs before the specified
 # branch.  So instead, export to a local tarball which is always found.
 # Make sure debhelper and dh-apport packages are installed before using this.
-PACKAGING := $(CURDIR)/../packaging.trunk
-PACKAGING_BRANCH := lp:~maas-maintainers/maas/packaging
+PACKAGING := $(CURDIR)/../packaging.utopic
+PACKAGING_BRANCH := lp:~maas-maintainers/maas/packaging.utopic
 
 package_branch:
 	@echo Downloading/refreshing packaging branch...

=== modified file 'docs/changelog.rst'
--- docs/changelog.rst	2014-11-05 03:36:42 +0000
+++ docs/changelog.rst	2014-11-06 01:24:10 +0000
@@ -2,472 +2,943 @@
 Changelog
 =========
 
-1.7.0
-=====
-
-Important announcements
------------------------
-
-**Re-import your boot images**
- You must re-import your boot images, see below for details.
-
-**Update Curtin preseed files**
- Two changes were made to Curtin preseed files that need your attention
- if you made any customisations:
-
- *  The OS name must now appear in the filename.  The new schema is shown
-    here, each file pattern is tried in turn until a match is found::
-
-    {prefix}_{osystem}_{node_arch}_{node_subarch}_{release}_{node_name}
-    {prefix}_{osystem}_{node_arch}_{node_subarch}_{release}
-    {prefix}_{osystem}_{node_arch}_{node_subarch}
-    {prefix}_{osystem}_{node_arch}
-    {prefix}_{osystem}
-    {prefix}
-
- * If you are modifying ``/etc/network/interfaces`` in the preseed, it must be
-   moved so it is processed last in ``late_commands`` since MAAS now writes
-   to this file itself as part of IPv6 setup.  For example::
-
-    late_commands:
-      bonding_02: ["curtin", "in-target", "--", "wget", "-O", "/etc/network/interfaces", "http://[...snip...]"]
-
-   must now look like this::
-
-    late_commands:
-      zz_write_ifaces: ["curtin", "in-target", "--", "wget", "-O", "/etc/network/interfaces", "http://[...snip...]"]
-
-   The leading ``zz`` ensures the command sorts to the end of the
-   ``late_commands`` list.
-
-
-Major new features
-------------------
-
-**Improved image downloading and reporting.**
-  MAAS boot images are now downloaded centrally by the region controller
-  and disseminated to all registered cluster controllers.  This change includes
-  a new web UI under the `Images` tab that allows the admin to select
-  which images to import and shows the progress of the ongoing download.
-  This completely replaces any file-based configuration that used to take
-  place on cluster controllers.  The cluster page now shows whether it has
-  synchronised all the images from the region controller.
-
-  This process is also completely controllable using the API.
-
-.. Note::
-  Unfortunately due to a format change in the way images are stored, it
-  was not possible to migrate previously downloaded images to the new region
-  storage.  The cluster(s) will still be able to use the existing images,
-  however the region controller will be unaware of them until an import
-  is initiated.  When the import is finished, the cluster(s) will remove
-  older image resources.
-
-  This means that the first thing to do after upgrading to 1.7 is go to the
-  `Images` tab and re-import the images.
-
-**Increased robustness.**
-  A large amount of effort has been given to ensuring that MAAS remains
-  robust in the face of adversity.  An updated node state model has been
-  implemented that takes into account more of the situations in which a
-  node can be found including any failures at each stage.
-
-  When a node is getting deployed, it is now monitored to check that each
-  stage is reached in a timely fashion; if it does not then it is marked
-  as failed.
-
-  The core power driver was updated to check the state of the power on each
-  node and is reported in the web UI and API.  The core driver now also
-  handles retries when changing the power state of hardware, removing the
-  requirement that each power template handle it individually.
-
-**RPC security.**
-  As a step towards mutually verified TLS connections between MAAS's
-  components, 1.7 introduces a simple shared-secret mechanism to
-  authenticate the region with the clusters and vice-versa. For those
-  clusters that run on the same machine as the region controller (which
-  will account for most people), everything will continue to work
-  without intervention. However, if you're running a cluster on a
-  separate machine, you must install the secret:
-
-  1. After upgrading the region controller, view /var/lib/maas/secret
-     (it's text) and copy it.
-
-  2. On each cluster, run:
-
-       sudo -u maas maas-provision install-shared-secret
-
-     You'll be prompted for the secret; paste it in and press enter. It
-     is a password prompt, so the secret will not be echoed back to you.
-
-  That's it; the upgraded cluster controller will find the secret
-  without needing to be told.
-
-**RPC connections.**
-  Each cluster maintains a persistent connection to each region
-  controller process that's running. The ports on which the region is
-  listening are all high-numbered, and they are allocated randomly by
-  the OS. In a future release of MAAS we will narrow this down. For now,
-  each cluster controller needs unfiltered access to each machine in the
-  region on all high-numbered TCP ports.
-
-**Node event log.**
-  For every major event on nodes, it is now logged in a node-specific log.
-  This includes events such as power changes, deployments and any failures.
-
-**IPv6.**
-  It is now possible to deploy Ubuntu nodes that have IPv6 enabled.
-  See :doc:`ipv6` for more details.
-
-**Removal of Celery and RabbitMQ.**
-  While Celery was found to be very reliable it ultimately did not suit
-  the project's requirements as it is a largely fire-and-forget mechanism.
-  Additionally it was another moving part that caused some headaches for
-  users and admins alike, so the decision was taken to remove it and implement
-  a custom communications mechanism between the region controller and cluster
-  controllers.  The new mechanism is bidirectional and allowed the complex
-  interactions to take place that are required as part of the robustness
-  improvements.
-
-  Since a constant connection is maintained, as a side effect the web UI now
-  shows whether each cluster is connected or not.
-
-**Support for other OSes.**
-  Non-Ubuntu OSes are fully supported now. This includes:
-   - Windows
-   - Centos
-   - SuSE
-
-**Custom Images.**
-  MAAS now supports the deployment of Custom Images. Custom images can be
-  uploaded via the API. The usage of custom images allows the deployment of
-  other Ubuntu Flavors, such as Ubuntu Desktop.
-
-**maas-proxy.**
-  MAAS now uses maas-proxy as the default proxy solution instead of
-  squid-deb-proxy. On a fresh install, MAAS will use maas-proxy by default.
-  On upgrades from previous releases, MAAS will install maas-proxy instead of
-  squid-deb-proxy.
-
-Minor notable changes
----------------------
-
-**Better handling of networks.**
-  All networks referred to by cluster interfaces are now automatically
-  registered on the Network page.  Any node network interfaces are
-  automatically linked to the relevant Network.
-
-.. Note::
-  Commissioning currently requires an IP address to be available for each
-  network interface on a network that MAAS manages; this allows MAAS to
-  auto-populate its networks database.  In general you should use a
-  well-sized network (/16 recommended if you will be using containers and
-  VMs) and dynamic pool. If this feature risks causing IP exhaustion for
-  your deployment and you do not need the auto-populate functionality, you
-  can disable it by running the following command on your region controller::
-
-    sudo maas <profile> maas set-config name=enable_dhcp_discovery_on_unconfigured_interfaces value=False
-
-**Improved logging.**
-  A total overhaul of where logging is produced was undertaken, and now
-  all the main events in MAAS are selectively reported to syslog with the
-  "maas" prefix from both the region and cluster controllers alike.  If MAAS
-  is installed using the standard Ubuntu packaging, its syslog entries are
-  redirected to /var/log/maas/maas.log.
-
-  On the clusters, pserv.log is now less chatty and contains only errors.
-  On the region controller appservers, maas-django.log contains only appserver
-  errors.
-
-**Static IP selection.**
- The API was extended so that specific IPs can be pre-allocated for network
- interfaces on nodes and for user-allocated IPs.
-
-**Pronounceable random hostnames.**
- The old auto-generated 5-letter names were replaced with a pseudo-random
- name that is produced from a dictionary giving names of the form
- 'adjective-noun'.
-
-
-Known Problems & Workarounds
-----------------------------
-
-**Upgrade issues**
- There may be upgrade issues for users currently on MAAS 1.5 and 1.6; while we
- have attempted to reproduce and address all the issues reported, some bugs
- remain inconclusive. We recommend a full, tested backup of the MAAS servers
- before attempting the upgrade to 1.7. If you do encounter issues, please file
- these and flag them to the attention of the MAAS team and we will address them
- in point-releases.  See bugs `1381058`_, `1382266`_, `1379890`_, `1379532`_,
- and `1379144`_.
-
-.. _1381058:
-  https://launchpad.net/bugs/1381058
-.. _1382266:
-  https://launchpad.net/bugs/1382266
-.. _1379890:
-  https://launchpad.net/bugs/1379890
-.. _1379532:
-  https://launchpad.net/bugs/1379532
-.. _1379144:
-  https://launchpad.net/bugs/1379144
-
-**Split Region/Cluster set-ups**
- If you site your cluster on a separate host to the region, it needs a
- security key to be manually installed by running
- ``maas-provision install-shared-secret`` on the cluster host.
-
-**Private boot streams**
- If you had private boot image stream information configured in MAAS 1.5 or
- 1.6, upgrading to 1.7 will not take that into account and it will need to be
- manually entered on the settings page in the MAAS UI (bug `1379890`_)
-
-.. _1379890:
-  https://launchpad.net/bugs/1379890
-
-**Concurrency issues**
- Concurrency issues expose us to races when simultaneous operations are
- triggered. This is the source of many hard to reproduce issues which will
- require us to change the default database isolation level. We intend to address
- this in the first point release of 1.7.
-
-**Destroying a Juju environment**
- When attempting to "juju destroy" an environment, nodes must be in the DEPLOYED
- state; otherwise, the destroy will fail. You should wait for all in-progress
- actions on the MAAS cluster to conclude before issuing the command. (bug
- `1381619`_)
-
-.. _1381619:
-  https://launchpad.net/bugs/_1381619
-
-**AMT power control**
- A few AMT-related issues remain, with workarounds:
-
-  * Commissioning NUC reboots instead of shutting down (bug `1368685`_).  There
-    is `a workaround in the power template`_
-
-  * MAAS (amttool) cannot control AMT version > 8. See `workaround described in
-    bug 1331214`_
-
-  * AMT NUC stuck at boot prompt instead of powering down (no ACPI support in
-    syslinux poweroff) (bug `1376716`_). See the `ACPI-only workaround`_
-
-.. _1368685:
-  https://bugs.launchpad.net/maas/+bug/1368685
-.. _a workaround in the power template:
-  https://bugs.launchpad.net/maas/+bug/1368685/comments/8
-.. _workaround described in bug 1331214:
-  https://bugs.launchpad.net/maas/+bug/1331214/comments/18
-.. _1376716:
-  https://bugs.launchpad.net/maas/+bug/1376716
-.. _ACPI-only workaround:
-  https://bugs.launchpad.net/maas/+bug/1376716/comments/12
-
-
-**Disk wiping**
- If you enable disk wiping, juju destroy-environment may fail for you. The
- current workaround is to wait and re-issue the command.  This will be fixed in
- future versions of MAAS & Juju. (bug `1386327`_)
-
-.. _1386327:
-  https://bugs.launchpad.net/maas/+bug/1386327
-
-**BIND with DNSSEC**
- If you are using BIND with a forwarder that uses DNSSEC and have not
- configured certificates, you will need to explicitly disable that feature in
- your BIND configuration (1384334)
-
-.. _1384334:
-  https://bugs.launchpad.net/maas/+bug/1384334
-
-**Boot source selections on the API**
- Use of API to change image selections can leave DB in a bad state
- (bug `1376812`_).  It can be fixed by issuing direct database updates.
-
-.. _1376812:
-  https://bugs.launchpad.net/maas/+bug/1376812
-
-**Disabling DNS**
- Disabling DNS may not work (bug `1383768`_)
-
-.. _1383768:
-  https://bugs.launchpad.net/maas/+bug/1383768
-
-**Stale DNS zone files**
- Stale DNS zone files may be left behind if the MAAS domainname is changed
- (bug `1383329`_)
-
-.. _1383329:
-  https://bugs.launchpad.net/maas/+bug/1383329
-
-
-
-Major bugs fixed in this release
---------------------------------
-
-See https://launchpad.net/maas/+milestone/1.7.0 for full details.
-
-#1081660    If maas-enlist fails to reach a DNS server, the node will be named ";; connection timed out; no servers could be reached"
-
-#1087183    MaaS cloud-init configuration specifies 'manage_etc_hosts: localhost'
-
-#1328351    ConstipationError: When the cluster runs the "import boot images" task it blocks other tasks
-
-#1342117    CLI command to set up node-group-interface fails with /usr/lib/python2.7/dist-packages/maascli/__main__.py: error: u'name'
-
-#1349254    Duplicate FQDN can be configured on MAAS via CLI or API
-
-#1352575    BMC password showing in the apache2 logs
-
-#1355534    UnknownPowerType traceback in appserver log
-
-#1363850    Auto-enlistment not reporting power parameters
-
-#1363900    Dev server errors while trying to write to '/var/lib/maas'
-
-#1363999    Not assigning static IP addresses
-
-#1364481    http 500 error doesn't contain a stack trace
-
-#1364993    500 error when trying to acquire a commissioned node (AddrFormatError: failed to detect a valid IP address from None)
-
-#1365130    django-admin prints spurious messages to stdout, breaking scripts
-
-#1365850    DHCP scan using cluster interface name as network interface?
-
-#1366172    NUC does not boot after power off/power on
-
-#1366212    Large dhcp leases file leads to tftp timeouts
-
-#1366652    Leaking temporary directories
-
-#1368269    internal server error when deleting a node
-
-#1368590    Power actions are not serialized.
-
-#1370534    Recurrent update of the power state of nodes crashes if the connection to the BMC fails.
-
-#1370958    excessive pserv logging
-
-#1372767    Twisted web client does not support IPv6 address
-
-#1372944    Twisted web client fails looking up IPv6 address hostname
-
-#1373031    Cannot register cluster
-
-#1373103    compose_curtin_network_preseed breaks installation of all other operating systems
-
-#1373368    Conflicting power actions being dropped on the floor can result in leaving a node in an inconsistent state
-
-#1373699    Cluster Listing Page lacks feedback about the images each cluster has
-
-#1374102    No retries for AMT power?
-
-#1375980    Nodes failed to transition out of "New" state on bulk commission
-
-#1376023    After performing bulk action on maas nodes, Internal Server Error
-
-#1376888    Nodes can't be deleted if DHCP management is off.
-
-#1377099    Bulk operation leaves nodes in inconsistent state
-
-#1379209    When a node has multiple interfaces on a network MAAS manages, MAAS assigns static IP addresses to all of them
-
-#1379744    Cluster registration is fragile and insecure
-
-#1380932    MAAS does not cope with changes of the dhcp daemons
-
-#1381605    Not all the DNS records are being added when deploying multiple nodes
-
-#1012954    If a power script fails, there is no UI feedback
-
-#1186196    "Starting a node" has different meanings in the UI and in the API.
-
-#1237215    maas and curtin do not indicate failure reasonably
-
-#1273222    MAAS doesn't check return values of power actions
-
-#1288502    archive and proxy settings not honoured for commissioning
-
-#1316919    Checks don't exist to confirm a node will actually boot
-
-#1321885    IPMI detection and automatic setting fail in ubuntu 14.04 maas
-
-#1325610    node marked "Ready" before poweroff complete
-
-#1340188    unallocated node started manually, causes AssertionError for purpose poweroff
-
-#1341118    No feedback when IPMI credentials fail
-
-#1341121    No feedback to user when cluster is not running
-
-#1341581    power state is not represented in api and ui
-
-#1341800    MAAS doesn't support soft power off through the API
-
-#1344177    hostnames can't be changed while a node is acquired
-
-#1347518    Confusing error message when API key is wrong
-
-#1349496    Unable to request a specific static IP on the API
-
-#1349736    MAAS logging is too verbose and not very useful
-
-#1349917    guess_server_address() can return IPAddress or hostname
-
-#1350103    No support for armhf/keystone architecture
-
-#1350856    Can't constrain acquisition of nodes by not having a tag
-
-#1356880    MAAS shouldn't allow changing the hostname of a deployed node
-
-#1357714    Virsh power driver does not seem to work at all
-
-#1358859    Commissioning output xml is hard to understand, would be nice to have yaml as an output option.
-
-#1359169    MAAS should handle invalid consumers gracefully
-
-#1359822    Gateway is missing in network definition
-
-#1363913    Impossible to remove last MAC from network in UI
-
-#1364228    Help text for node hostname is wrong
-
-#1364591    MAAS Archive Mirror does not respect non-default port
-
-#1365616    Non-admin access to cluster controller config
-
-#1365619    DNS should be an optional field in the network definition
-
-#1365776    commissioning results view for a node also shows installation results
-
-#1366812    Old boot resources are not being removed on clusters
-
-#1367455    MAC address for node's IPMI is reversed looked up to yield IP address using case sensitive comparison
-
-#1373580    [SRU] Glen m700 cartridge list as ARM64/generic after enlist
-
-#1373723    Releasing a node without power parameters ends up in not being able to release a node
-
-#1233158    no way to get power parameters in api
-
-#1319854    `maas login` tells you you're logged in successfully when you're not
-
-#1368480    Need API to gather image metadata across all of MAAS
-
-#1281406    Disk/memory space on Node edit page have no units
-
-#1299231    MAAS DHCP/DNS can't manage more than a /16 network
-
-#1357381    maas-region-admin createadmin shows error if not params given
-
-#1376393    powerkvm boot loader installs even when not needed
-
-#1287224    MAAS random generated hostnames are not pronounceable
-
-#1348364    non-maas managed subnets cannot query maas DNS
-
-
+<<<<<<< TREE
+1.7.0
+=====
+
+Important announcements
+-----------------------
+
+**Re-import your boot images**
+ You must re-import your boot images, see below for details.
+
+**Update Curtin preseed files**
+ Two changes were made to Curtin preseed files that need your attention
+ if you made any customisations:
+
+ *  The OS name must now appear in the filename.  The new schema is shown
+    here, each file pattern is tried in turn until a match is found::
+
+    {prefix}_{osystem}_{node_arch}_{node_subarch}_{release}_{node_name}
+    {prefix}_{osystem}_{node_arch}_{node_subarch}_{release}
+    {prefix}_{osystem}_{node_arch}_{node_subarch}
+    {prefix}_{osystem}_{node_arch}
+    {prefix}_{osystem}
+    {prefix}
+
+ * If you are modifying ``/etc/network/interfaces`` in the preseed, it must be
+   moved so it is processed last in ``late_commands`` since MAAS now writes
+   to this file itself as part of IPv6 setup.  For example::
+
+    late_commands:
+      bonding_02: ["curtin", "in-target", "--", "wget", "-O", "/etc/network/interfaces", "http://[...snip...]"]
+
+   must now look like this::
+
+    late_commands:
+      zz_write_ifaces: ["curtin", "in-target", "--", "wget", "-O", "/etc/network/interfaces", "http://[...snip...]"]
+
+   The leading ``zz`` ensures the command sorts to the end of the
+   ``late_commands`` list.
+
+
+Major new features
+------------------
+
+**Improved image downloading and reporting.**
+  MAAS boot images are now downloaded centrally by the region controller
+  and disseminated to all registered cluster controllers.  This change includes
+  a new web UI under the `Images` tab that allows the admin to select
+  which images to import and shows the progress of the ongoing download.
+  This completely replaces any file-based configuration that used to take
+  place on cluster controllers.  The cluster page now shows whether it has
+  synchronised all the images from the region controller.
+
+  This process is also completely controllable using the API.
+
+.. Note::
+  Unfortunately due to a format change in the way images are stored, it
+  was not possible to migrate previously downloaded images to the new region
+  storage.  The cluster(s) will still be able to use the existing images,
+  however the region controller will be unaware of them until an import
+  is initiated.  When the import is finished, the cluster(s) will remove
+  older image resources.
+
+  This means that the first thing to do after upgrading to 1.7 is go to the
+  `Images` tab and re-import the images.
+
+**Increased robustness.**
+  A large amount of effort has been given to ensuring that MAAS remains
+  robust in the face of adversity.  An updated node state model has been
+  implemented that takes into account more of the situations in which a
+  node can be found including any failures at each stage.
+
+  When a node is getting deployed, it is now monitored to check that each
+  stage is reached in a timely fashion; if it does not then it is marked
+  as failed.
+
+  The core power driver was updated to check the state of the power on each
+  node and is reported in the web UI and API.  The core driver now also
+  handles retries when changing the power state of hardware, removing the
+  requirement that each power template handle it individually.
+
+**RPC security.**
+  As a step towards mutually verified TLS connections between MAAS's
+  components, 1.7 introduces a simple shared-secret mechanism to
+  authenticate the region with the clusters and vice-versa. For those
+  clusters that run on the same machine as the region controller (which
+  will account for most people), everything will continue to work
+  without intervention. However, if you're running a cluster on a
+  separate machine, you must install the secret:
+
+  1. After upgrading the region controller, view /var/lib/maas/secret
+     (it's text) and copy it.
+
+  2. On each cluster, run:
+
+       sudo -u maas maas-provision install-shared-secret
+
+     You'll be prompted for the secret; paste it in and press enter. It
+     is a password prompt, so the secret will not be echoed back to you.
+
+  That's it; the upgraded cluster controller will find the secret
+  without needing to be told.
+
+**RPC connections.**
+  Each cluster maintains a persistent connection to each region
+  controller process that's running. The ports on which the region is
+  listening are all high-numbered, and they are allocated randomly by
+  the OS. In a future release of MAAS we will narrow this down. For now,
+  each cluster controller needs unfiltered access to each machine in the
+  region on all high-numbered TCP ports.
+
+**Node event log.**
+  For every major event on nodes, it is now logged in a node-specific log.
+  This includes events such as power changes, deployments and any failures.
+
+**IPv6.**
+  It is now possible to deploy Ubuntu nodes that have IPv6 enabled.
+  See :doc:`ipv6` for more details.
+
+**Removal of Celery and RabbitMQ.**
+  While Celery was found to be very reliable it ultimately did not suit
+  the project's requirements as it is a largely fire-and-forget mechanism.
+  Additionally it was another moving part that caused some headaches for
+  users and admins alike, so the decision was taken to remove it and implement
+  a custom communications mechanism between the region controller and cluster
+  controllers.  The new mechanism is bidirectional and allowed the complex
+  interactions to take place that are required as part of the robustness
+  improvements.
+
+  Since a constant connection is maintained, as a side effect the web UI now
+  shows whether each cluster is connected or not.
+
+**Support for other OSes.**
+  Non-Ubuntu OSes are fully supported now. This includes:
+   - Windows
+   - Centos
+   - SuSE
+
+**Custom Images.**
+  MAAS now supports the deployment of Custom Images. Custom images can be
+  uploaded via the API. The usage of custom images allows the deployment of
+  other Ubuntu Flavors, such as Ubuntu Desktop.
+
+**maas-proxy.**
+  MAAS now uses maas-proxy as the default proxy solution instead of
+  squid-deb-proxy. On a fresh install, MAAS will use maas-proxy by default.
+  On upgrades from previous releases, MAAS will install maas-proxy instead of
+  squid-deb-proxy.
+
+Minor notable changes
+---------------------
+
+**Better handling of networks.**
+  All networks referred to by cluster interfaces are now automatically
+  registered on the Network page.  Any node network interfaces are
+  automatically linked to the relevant Network.
+
+.. Note::
+  Commissioning currently requires an IP address to be available for each
+  network interface on a network that MAAS manages; this allows MAAS to
+  auto-populate its networks database.  In general you should use a
+  well-sized network (/16 recommended if you will be using containers and
+  VMs) and dynamic pool. If this feature risks causing IP exhaustion for
+  your deployment and you do not need the auto-populate functionality, you
+  can disable it by running the following command on your region controller::
+
+    sudo maas <profile> maas set-config name=enable_dhcp_discovery_on_unconfigured_interfaces value=False
+
+**Improved logging.**
+  A total overhaul of where logging is produced was undertaken, and now
+  all the main events in MAAS are selectively reported to syslog with the
+  "maas" prefix from both the region and cluster controllers alike.  If MAAS
+  is installed using the standard Ubuntu packaging, its syslog entries are
+  redirected to /var/log/maas/maas.log.
+
+  On the clusters, pserv.log is now less chatty and contains only errors.
+  On the region controller appservers, maas-django.log contains only appserver
+  errors.
+
+**Static IP selection.**
+ The API was extended so that specific IPs can be pre-allocated for network
+ interfaces on nodes and for user-allocated IPs.
+
+**Pronounceable random hostnames.**
+ The old auto-generated 5-letter names were replaced with a pseudo-random
+ name that is produced from a dictionary giving names of the form
+ 'adjective-noun'.
+
+
+Known Problems & Workarounds
+----------------------------
+
+**Upgrade issues**
+ There may be upgrade issues for users currently on MAAS 1.5 and 1.6; while we
+ have attempted to reproduce and address all the issues reported, some bugs
+ remain inconclusive. We recommend a full, tested backup of the MAAS servers
+ before attempting the upgrade to 1.7. If you do encounter issues, please file
+ these and flag them to the attention of the MAAS team and we will address them
+ in point-releases.  See bugs `1381058`_, `1382266`_, `1379890`_, `1379532`_,
+ and `1379144`_.
+
+.. _1381058:
+  https://launchpad.net/bugs/1381058
+.. _1382266:
+  https://launchpad.net/bugs/1382266
+.. _1379890:
+  https://launchpad.net/bugs/1379890
+.. _1379532:
+  https://launchpad.net/bugs/1379532
+.. _1379144:
+  https://launchpad.net/bugs/1379144
+
+**Split Region/Cluster set-ups**
+ If you site your cluster on a separate host to the region, it needs a
+ security key to be manually installed by running
+ ``maas-provision install-shared-secret`` on the cluster host.
+
+**Private boot streams**
+ If you had private boot image stream information configured in MAAS 1.5 or
+ 1.6, upgrading to 1.7 will not take that into account and it will need to be
+ manually entered on the settings page in the MAAS UI (bug `1379890`_)
+
+.. _1379890:
+  https://launchpad.net/bugs/1379890
+
+**Concurrency issues**
+ Concurrency issues expose us to races when simultaneous operations are
+ triggered. This is the source of many hard to reproduce issues which will
+ require us to change the default database isolation level. We intend to address
+ this in the first point release of 1.7.
+
+**Destroying a Juju environment**
+ When attempting to "juju destroy" an environment, nodes must be in the DEPLOYED
+ state; otherwise, the destroy will fail. You should wait for all in-progress
+ actions on the MAAS cluster to conclude before issuing the command. (bug
+ `1381619`_)
+
+.. _1381619:
+  https://launchpad.net/bugs/_1381619
+
+**AMT power control**
+ A few AMT-related issues remain, with workarounds:
+
+  * Commissioning NUC reboots instead of shutting down (bug `1368685`_).  There
+    is `a workaround in the power template`_
+
+  * MAAS (amttool) cannot control AMT version > 8. See `workaround described in
+    bug 1331214`_
+
+  * AMT NUC stuck at boot prompt instead of powering down (no ACPI support in
+    syslinux poweroff) (bug `1376716`_). See the `ACPI-only workaround`_
+
+.. _1368685:
+  https://bugs.launchpad.net/maas/+bug/1368685
+.. _a workaround in the power template:
+  https://bugs.launchpad.net/maas/+bug/1368685/comments/8
+.. _workaround described in bug 1331214:
+  https://bugs.launchpad.net/maas/+bug/1331214/comments/18
+.. _1376716:
+  https://bugs.launchpad.net/maas/+bug/1376716
+.. _ACPI-only workaround:
+  https://bugs.launchpad.net/maas/+bug/1376716/comments/12
+
+
+**Disk wiping**
+ If you enable disk wiping, juju destroy-environment may fail for you. The
+ current workaround is to wait and re-issue the command.  This will be fixed in
+ future versions of MAAS & Juju. (bug `1386327`_)
+
+.. _1386327:
+  https://bugs.launchpad.net/maas/+bug/1386327
+
+**BIND with DNSSEC**
+ If you are using BIND with a forwarder that uses DNSSEC and have not
+ configured certificates, you will need to explicitly disable that feature in
+ your BIND configuration (1384334)
+
+.. _1384334:
+  https://bugs.launchpad.net/maas/+bug/1384334
+
+**Boot source selections on the API**
+ Use of API to change image selections can leave DB in a bad state
+ (bug `1376812`_).  It can be fixed by issuing direct database updates.
+
+.. _1376812:
+  https://bugs.launchpad.net/maas/+bug/1376812
+
+**Disabling DNS**
+ Disabling DNS may not work (bug `1383768`_)
+
+.. _1383768:
+  https://bugs.launchpad.net/maas/+bug/1383768
+
+**Stale DNS zone files**
+ Stale DNS zone files may be left behind if the MAAS domainname is changed
+ (bug `1383329`_)
+
+.. _1383329:
+  https://bugs.launchpad.net/maas/+bug/1383329
+
+
+
+Major bugs fixed in this release
+--------------------------------
+
+See https://launchpad.net/maas/+milestone/1.7.0 for full details.
+
+#1081660    If maas-enlist fails to reach a DNS server, the node will be named ";; connection timed out; no servers could be reached"
+
+#1087183    MaaS cloud-init configuration specifies 'manage_etc_hosts: localhost'
+
+#1328351    ConstipationError: When the cluster runs the "import boot images" task it blocks other tasks
+
+#1342117    CLI command to set up node-group-interface fails with /usr/lib/python2.7/dist-packages/maascli/__main__.py: error: u'name'
+
+#1349254    Duplicate FQDN can be configured on MAAS via CLI or API
+
+#1352575    BMC password showing in the apache2 logs
+
+#1355534    UnknownPowerType traceback in appserver log
+
+#1363850    Auto-enlistment not reporting power parameters
+
+#1363900    Dev server errors while trying to write to '/var/lib/maas'
+
+#1363999    Not assigning static IP addresses
+
+#1364481    http 500 error doesn't contain a stack trace
+
+#1364993    500 error when trying to acquire a commissioned node (AddrFormatError: failed to detect a valid IP address from None)
+
+#1365130    django-admin prints spurious messages to stdout, breaking scripts
+
+#1365850    DHCP scan using cluster interface name as network interface?
+
+#1366172    NUC does not boot after power off/power on
+
+#1366212    Large dhcp leases file leads to tftp timeouts
+
+#1366652    Leaking temporary directories
+
+#1368269    internal server error when deleting a node
+
+#1368590    Power actions are not serialized.
+
+#1370534    Recurrent update of the power state of nodes crashes if the connection to the BMC fails.
+
+#1370958    excessive pserv logging
+
+#1372767    Twisted web client does not support IPv6 address
+
+#1372944    Twisted web client fails looking up IPv6 address hostname
+
+#1373031    Cannot register cluster
+
+#1373103    compose_curtin_network_preseed breaks installation of all other operating systems
+
+#1373368    Conflicting power actions being dropped on the floor can result in leaving a node in an inconsistent state
+
+#1373699    Cluster Listing Page lacks feedback about the images each cluster has
+
+#1374102    No retries for AMT power?
+
+#1375980    Nodes failed to transition out of "New" state on bulk commission
+
+#1376023    After performing bulk action on maas nodes, Internal Server Error
+
+#1376888    Nodes can't be deleted if DHCP management is off.
+
+#1377099    Bulk operation leaves nodes in inconsistent state
+
+#1379209    When a node has multiple interfaces on a network MAAS manages, MAAS assigns static IP addresses to all of them
+
+#1379744    Cluster registration is fragile and insecure
+
+#1380932    MAAS does not cope with changes of the dhcp daemons
+
+#1381605    Not all the DNS records are being added when deploying multiple nodes
+
+#1012954    If a power script fails, there is no UI feedback
+
+#1186196    "Starting a node" has different meanings in the UI and in the API.
+
+#1237215    maas and curtin do not indicate failure reasonably
+
+#1273222    MAAS doesn't check return values of power actions
+
+#1288502    archive and proxy settings not honoured for commissioning
+
+#1316919    Checks don't exist to confirm a node will actually boot
+
+#1321885    IPMI detection and automatic setting fail in ubuntu 14.04 maas
+
+#1325610    node marked "Ready" before poweroff complete
+
+#1340188    unallocated node started manually, causes AssertionError for purpose poweroff
+
+#1341118    No feedback when IPMI credentials fail
+
+#1341121    No feedback to user when cluster is not running
+
+#1341581    power state is not represented in api and ui
+
+#1341800    MAAS doesn't support soft power off through the API
+
+#1344177    hostnames can't be changed while a node is acquired
+
+#1347518    Confusing error message when API key is wrong
+
+#1349496    Unable to request a specific static IP on the API
+
+#1349736    MAAS logging is too verbose and not very useful
+
+#1349917    guess_server_address() can return IPAddress or hostname
+
+#1350103    No support for armhf/keystone architecture
+
+#1350856    Can't constrain acquisition of nodes by not having a tag
+
+#1356880    MAAS shouldn't allow changing the hostname of a deployed node
+
+#1357714    Virsh power driver does not seem to work at all
+
+#1358859    Commissioning output xml is hard to understand, would be nice to have yaml as an output option.
+
+#1359169    MAAS should handle invalid consumers gracefully
+
+#1359822    Gateway is missing in network definition
+
+#1363913    Impossible to remove last MAC from network in UI
+
+#1364228    Help text for node hostname is wrong
+
+#1364591    MAAS Archive Mirror does not respect non-default port
+
+#1365616    Non-admin access to cluster controller config
+
+#1365619    DNS should be an optional field in the network definition
+
+#1365776    commissioning results view for a node also shows installation results
+
+#1366812    Old boot resources are not being removed on clusters
+
+#1367455    MAC address for node's IPMI is reversed looked up to yield IP address using case sensitive comparison
+
+#1373580    [SRU] Glen m700 cartridge list as ARM64/generic after enlist
+
+#1373723    Releasing a node without power parameters ends up in not being able to release a node
+
+#1233158    no way to get power parameters in api
+
+#1319854    `maas login` tells you you're logged in successfully when you're not
+
+#1368480    Need API to gather image metadata across all of MAAS
+
+#1281406    Disk/memory space on Node edit page have no units
+
+#1299231    MAAS DHCP/DNS can't manage more than a /16 network
+
+#1357381    maas-region-admin createadmin shows error if not params given
+
+#1376393    powerkvm boot loader installs even when not needed
+
+#1287224    MAAS random generated hostnames are not pronounceable
+
+#1348364    non-maas managed subnets cannot query maas DNS
+
+
+=======
+1.7.0
+=====
+
+Important announcements
+-----------------------
+
+**Re-import your boot images**
+ You must re-import your boot images, see below for details.
+
+**Update Curtin preseed files**
+ Two changes were made to Curtin preseed files that need your attention
+ if you made any customisations:
+
+ *  The OS name must now appear in the filename.  The new schema is shown
+    here, each file pattern is tried in turn until a match is found::
+
+    {prefix}_{osystem}_{node_arch}_{node_subarch}_{release}_{node_name}
+    {prefix}_{osystem}_{node_arch}_{node_subarch}_{release}
+    {prefix}_{osystem}_{node_arch}_{node_subarch}
+    {prefix}_{osystem}_{node_arch}
+    {prefix}_{osystem}
+    {prefix}
+
+ * If you are modifying ``/etc/network/interfaces`` in the preseed, it must be
+   moved so it is processed last in ``late_commands`` since MAAS now writes
+   to this file itself as part of IPv6 setup.  For example::
+
+    late_commands:
+      bonding_02: ["curtin", "in-target", "--", "wget", "-O", "/etc/network/interfaces", "http://[...snip...]"]
+
+   must now look like this::
+
+    late_commands:
+      zz_write_ifaces: ["curtin", "in-target", "--", "wget", "-O", "/etc/network/interfaces", "http://[...snip...]"]
+
+   The leading ``zz`` ensures the command sorts to the end of the
+   ``late_commands`` list.
+
+
+Major new features
+------------------
+
+**Improved image downloading and reporting.**
+  MAAS boot images are now downloaded centrally by the region controller
+  and disseminated to all registered cluster controllers.  This change includes
+  a new web UI under the `Images` tab that allows the admin to select
+  which images to import and shows the progress of the ongoing download.
+  This completely replaces any file-based configuration that used to take
+  place on cluster controllers.  The cluster page now shows whether it has
+  synchronised all the images from the region controller.
+
+  This process is also completely controllable using the API.
+
+.. Note::
+  Unfortunately due to a format change in the way images are stored, it
+  was not possible to migrate previously downloaded images to the new region
+  storage.  The cluster(s) will still be able to use the existing images,
+  however the region controller will be unaware of them until an import
+  is initiated.  When the import is finished, the cluster(s) will remove
+  older image resources.
+
+  This means that the first thing to do after upgrading to 1.7 is go to the
+  `Images` tab and re-import the images.
+
+**Increased robustness.**
+  A large amount of effort has been given to ensuring that MAAS remains
+  robust in the face of adversity.  An updated node state model has been
+  implemented that takes into account more of the situations in which a
+  node can be found including any failures at each stage.
+
+  When a node is getting deployed, it is now monitored to check that each
+  stage is reached in a timely fashion; if it does not then it is marked
+  as failed.
+
+  The core power driver was updated to check the state of the power on each
+  node and is reported in the web UI and API.  The core driver now also
+  handles retries when changing the power state of hardware, removing the
+  requirement that each power template handle it individually.
+
+**RPC security.**
+  As a step towards mutually verified TLS connections between MAAS's
+  components, 1.7 introduces a simple shared-secret mechanism to
+  authenticate the region with the clusters and vice-versa. For those
+  clusters that run on the same machine as the region controller (which
+  will account for most people), everything will continue to work
+  without intervention. However, if you're running a cluster on a
+  separate machine, you must install the secret:
+
+  1. After upgrading the region controller, view /var/lib/maas/secret
+     (it's text) and copy it.
+
+  2. On each cluster, run:
+
+       sudo -u maas maas-provision install-shared-secret
+
+     You'll be prompted for the secret; paste it in and press enter. It
+     is a password prompt, so the secret will not be echoed back to you.
+
+  That's it; the upgraded cluster controller will find the secret
+  without needing to be told.
+
+**RPC connections.**
+  Each cluster maintains a persistent connection to each region
+  controller process that's running. The ports on which the region is
+  listening are all high-numbered, and they are allocated randomly by
+  the OS. In a future release of MAAS we will narrow this down. For now,
+  each cluster controller needs unfiltered access to each machine in the
+  region on all high-numbered TCP ports.
+
+**Node event log.**
+  For every major event on nodes, it is now logged in a node-specific log.
+  This includes events such as power changes, deployments and any failures.
+
+**IPv6.**
+  It is now possible to deploy Ubuntu nodes that have IPv6 enabled.
+  See :doc:`ipv6` for more details.
+
+**Removal of Celery and RabbitMQ.**
+  While Celery was found to be very reliable it ultimately did not suit
+  the project's requirements as it is a largely fire-and-forget mechanism.
+  Additionally it was another moving part that caused some headaches for
+  users and admins alike, so the decision was taken to remove it and implement
+  a custom communications mechanism between the region controller and cluster
+  controllers.  The new mechanism is bidirectional and allowed the complex
+  interactions to take place that are required as part of the robustness
+  improvements.
+
+  Since a constant connection is maintained, as a side effect the web UI now
+  shows whether each cluster is connected or not.
+
+**Support for other OSes.**
+  Non-Ubuntu OSes are fully supported now. This includes:
+   - Windows
+   - Centos
+   - SuSE
+
+**Custom Images.**
+  MAAS now supports the deployment of Custom Images. Custom images can be
+  uploaded via the API. The usage of custom images allows the deployment of
+  other Ubuntu Flavors, such as Ubuntu Desktop.
+
+**maas-proxy.**
+  MAAS now uses maas-proxy as the default proxy solution instead of
+  squid-deb-proxy. On a fresh install, MAAS will use maas-proxy by default.
+  On upgrades from previous releases, MAAS will install maas-proxy instead of
+  squid-deb-proxy.
+
+Minor notable changes
+---------------------
+
+**Better handling of networks.**
+  All networks referred to by cluster interfaces are now automatically
+  registered on the Network page.  Any node network interfaces are
+  automatically linked to the relevant Network.
+
+.. Note::
+  Commissioning currently requires an IP address to be available for each
+  network interface on a network that MAAS manages; this allows MAAS to
+  auto-populate its networks database.  In general you should use a
+  well-sized network (/16 recommended if you will be using containers and
+  VMs) and dynamic pool. If this feature risks causing IP exhaustion for
+  your deployment and you do not need the auto-populate functionality, you
+  can disable it by running the following command on your region controller::
+
+    sudo maas <profile> maas set-config name=enable_dhcp_discovery_on_unconfigured_interfaces value=False
+
+**Improved logging.**
+  A total overhaul of where logging is produced was undertaken, and now
+  all the main events in MAAS are selectively reported to syslog with the
+  "maas" prefix from both the region and cluster controllers alike.  If MAAS
+  is installed using the standard Ubuntu packaging, its syslog entries are
+  redirected to /var/log/maas/maas.log.
+
+  On the clusters, pserv.log is now less chatty and contains only errors.
+  On the region controller appservers, maas-django.log contains only appserver
+  errors.
+
+**Static IP selection.**
+ The API was extended so that specific IPs can be pre-allocated for network
+ interfaces on nodes and for user-allocated IPs.
+
+**Pronounceable random hostnames.**
+ The old auto-generated 5-letter names were replaced with a pseudo-random
+ name that is produced from a dictionary giving names of the form
+ 'adjective-noun'.
+
+
+Known Problems & Workarounds
+----------------------------
+
+**Upgrade issues**
+ There may be upgrade issues for users currently on MAAS 1.5 and 1.6; while we
+ have attempted to reproduce and address all the issues reported, some bugs
+ remain inconclusive. We recommend a full, tested backup of the MAAS servers
+ before attempting the upgrade to 1.7. If you do encounter issues, please file
+ these and flag them to the attention of the MAAS team and we will address them
+ in point-releases.  See bugs `1381058`_, `1382266`_, `1379890`_, `1379532`_,
+ and `1379144`_.
+
+.. _1381058:
+  https://launchpad.net/bugs/1381058
+.. _1382266:
+  https://launchpad.net/bugs/1382266
+.. _1379890:
+  https://launchpad.net/bugs/1379890
+.. _1379532:
+  https://launchpad.net/bugs/1379532
+.. _1379144:
+  https://launchpad.net/bugs/1379144
+
+**Split Region/Cluster set-ups**
+ If you site your cluster on a separate host to the region, it needs a
+ security key to be manually installed by running
+ ``maas-provision install-shared-secret`` on the cluster host.
+
+**Private boot streams**
+ If you had private boot image stream information configured in MAAS 1.5 or
+ 1.6, upgrading to 1.7 will not take that into account and it will need to be
+ manually entered on the settings page in the MAAS UI (bug `1379890`_)
+
+.. _1379890:
+  https://launchpad.net/bugs/1379890
+
+**Concurrency issues**
+ Concurrency issues expose us to races when simultaneous operations are
+ triggered. This is the source of many hard to reproduce issues which will
+ require us to change the default database isolation level. We intend to address
+ this in the first point release of 1.7.
+
+**Destroying a Juju environment**
+ When attempting to "juju destroy" an environment, nodes must be in the DEPLOYED
+ state; otherwise, the destroy will fail. You should wait for all in-progress
+ actions on the MAAS cluster to conclude before issuing the command. (bug
+ `1381619`_)
+
+.. _1381619:
+  https://launchpad.net/bugs/_1381619
+
+**AMT power control**
+ A few AMT-related issues remain, with workarounds:
+
+  * Commissioning NUC reboots instead of shutting down (bug `1368685`_).  There
+    is `a workaround in the power template`_
+
+  * MAAS (amttool) cannot control AMT version > 8. See `workaround described in
+    bug 1331214`_
+
+  * AMT NUC stuck at boot prompt instead of powering down (no ACPI support in
+    syslinux poweroff) (bug `1376716`_). See the `ACPI-only workaround`_
+
+.. _1368685:
+  https://bugs.launchpad.net/maas/+bug/1368685
+.. _a workaround in the power template:
+  https://bugs.launchpad.net/maas/+bug/1368685/comments/8
+.. _workaround described in bug 1331214:
+  https://bugs.launchpad.net/maas/+bug/1331214/comments/18
+.. _1376716:
+  https://bugs.launchpad.net/maas/+bug/1376716
+.. _ACPI-only workaround:
+  https://bugs.launchpad.net/maas/+bug/1376716/comments/12
+
+
+**Disk wiping**
+ If you enable disk wiping, juju destroy-environment may fail for you. The
+ current workaround is to wait and re-issue the command.  This will be fixed in
+ future versions of MAAS & Juju. (bug `1386327`_)
+
+.. _1386327:
+  https://bugs.launchpad.net/maas/+bug/1386327
+
+**BIND with DNSSEC**
+ If you are using BIND with a forwarder that uses DNSSEC and have not
+ configured certificates, you will need to explicitly disable that feature in
+ your BIND configuration (1384334)
+
+.. _1384334:
+  https://bugs.launchpad.net/maas/+bug/1384334
+
+**Boot source selections on the API**
+ Use of API to change image selections can leave DB in a bad state
+ (bug `1376812`_).  It can be fixed by issuing direct database updates.
+
+.. _1376812:
+  https://bugs.launchpad.net/maas/+bug/1376812
+
+**Disabling DNS**
+ Disabling DNS may not work (bug `1383768`_)
+
+.. _1383768:
+  https://bugs.launchpad.net/maas/+bug/1383768
+
+**Stale DNS zone files**
+ Stale DNS zone files may be left behind if the MAAS domainname is changed
+ (bug `1383329`_)
+
+.. _1383329:
+  https://bugs.launchpad.net/maas/+bug/1383329
+
+
+
+Major bugs fixed in this release
+--------------------------------
+
+See https://launchpad.net/maas/+milestone/1.7.0 for full details.
+
+#1081660    If maas-enlist fails to reach a DNS server, the node will be named ";; connection timed out; no servers could be reached"
+
+#1087183    MaaS cloud-init configuration specifies 'manage_etc_hosts: localhost'
+
+#1328351    ConstipationError: When the cluster runs the "import boot images" task it blocks other tasks
+
+#1342117    CLI command to set up node-group-interface fails with /usr/lib/python2.7/dist-packages/maascli/__main__.py: error: u'name'
+
+#1349254    Duplicate FQDN can be configured on MAAS via CLI or API
+
+#1352575    BMC password showing in the apache2 logs
+
+#1355534    UnknownPowerType traceback in appserver log
+
+#1363850    Auto-enlistment not reporting power parameters
+
+#1363900    Dev server errors while trying to write to '/var/lib/maas'
+
+#1363999    Not assigning static IP addresses
+
+#1364481    http 500 error doesn't contain a stack trace
+
+#1364993    500 error when trying to acquire a commissioned node (AddrFormatError: failed to detect a valid IP address from None)
+
+#1365130    django-admin prints spurious messages to stdout, breaking scripts
+
+#1365850    DHCP scan using cluster interface name as network interface?
+
+#1366172    NUC does not boot after power off/power on
+
+#1366212    Large dhcp leases file leads to tftp timeouts
+
+#1366652    Leaking temporary directories
+
+#1368269    internal server error when deleting a node
+
+#1368590    Power actions are not serialized.
+
+#1370534    Recurrent update of the power state of nodes crashes if the connection to the BMC fails.
+
+#1370958    excessive pserv logging
+
+#1372767    Twisted web client does not support IPv6 address
+
+#1372944    Twisted web client fails looking up IPv6 address hostname
+
+#1373031    Cannot register cluster
+
+#1373103    compose_curtin_network_preseed breaks installation of all other operating systems
+
+#1373368    Conflicting power actions being dropped on the floor can result in leaving a node in an inconsistent state
+
+#1373699    Cluster Listing Page lacks feedback about the images each cluster has
+
+#1374102    No retries for AMT power?
+
+#1375980    Nodes failed to transition out of "New" state on bulk commission
+
+#1376023    After performing bulk action on maas nodes, Internal Server Error
+
+#1376888    Nodes can't be deleted if DHCP management is off.
+
+#1377099    Bulk operation leaves nodes in inconsistent state
+
+#1379209    When a node has multiple interfaces on a network MAAS manages, MAAS assigns static IP addresses to all of them
+
+#1379744    Cluster registration is fragile and insecure
+
+#1380932    MAAS does not cope with changes of the dhcp daemons
+
+#1381605    Not all the DNS records are being added when deploying multiple nodes
+
+#1012954    If a power script fails, there is no UI feedback
+
+#1186196    "Starting a node" has different meanings in the UI and in the API.
+
+#1237215    maas and curtin do not indicate failure reasonably
+
+#1273222    MAAS doesn't check return values of power actions
+
+#1288502    archive and proxy settings not honoured for commissioning
+
+#1316919    Checks don't exist to confirm a node will actually boot
+
+#1321885    IPMI detection and automatic setting fail in ubuntu 14.04 maas
+
+#1325610    node marked "Ready" before poweroff complete
+
+#1325638    Add hardware enablement for Universal Management Gateway
+
+#1340188    unallocated node started manually, causes AssertionError for purpose poweroff
+
+#1341118    No feedback when IPMI credentials fail
+
+#1341121    No feedback to user when cluster is not running
+
+#1341581    power state is not represented in api and ui
+
+#1341800    MAAS doesn't support soft power off through the API
+
+#1344177    hostnames can't be changed while a node is acquired
+
+#1347518    Confusing error message when API key is wrong
+
+#1349496    Unable to request a specific static IP on the API
+
+#1349736    MAAS logging is too verbose and not very useful
+
+#1349917    guess_server_address() can return IPAddress or hostname
+
+#1350103    No support for armhf/keystone architecture
+
+#1350856    Can't constrain acquisition of nodes by not having a tag
+
+#1356880    MAAS shouldn't allow changing the hostname of a deployed node
+
+#1357714    Virsh power driver does not seem to work at all
+
+#1358859    Commissioning output xml is hard to understand, would be nice to have yaml as an output option.
+
+#1359169    MAAS should handle invalid consumers gracefully
+
+#1359822    Gateway is missing in network definition
+
+#1363913    Impossible to remove last MAC from network in UI
+
+#1364228    Help text for node hostname is wrong
+
+#1364591    MAAS Archive Mirror does not respect non-default port
+
+#1365616    Non-admin access to cluster controller config
+
+#1365619    DNS should be an optional field in the network definition
+
+#1365776    commissioning results view for a node also shows installation results
+
+#1366812    Old boot resources are not being removed on clusters
+
+#1367455    MAC address for node's IPMI is reversed looked up to yield IP address using case sensitive comparison
+
+#1373580    [SRU] Glen m700 cartridge list as ARM64/generic after enlist
+
+#1373723    Releasing a node without power parameters ends up in not being able to release a node
+
+#1233158    no way to get power parameters in api
+
+#1319854    `maas login` tells you you're logged in successfully when you're not
+
+#1368480    Need API to gather image metadata across all of MAAS
+
+#1281406    Disk/memory space on Node edit page have no units
+
+#1299231    MAAS DHCP/DNS can't manage more than a /16 network
+
+#1357381    maas-region-admin createadmin shows error if not params given
+
+#1376393    powerkvm boot loader installs even when not needed
+
+#1287224    MAAS random generated hostnames are not pronounceable
+
+#1348364    non-maas managed subnets cannot query maas DNS
+
+
+>>>>>>> MERGE-SOURCE
 1.6.1
 =====
 

=== modified file 'src/maasserver/api/nodes.py'
=== modified file 'src/maasserver/api/tests/test_enlistment.py'
=== modified file 'src/maasserver/api/tests/test_node.py'
--- src/maasserver/api/tests/test_node.py	2014-11-05 16:29:35 +0000
+++ src/maasserver/api/tests/test_node.py	2014-11-06 01:24:10 +0000
@@ -26,7 +26,12 @@
 from maasserver.enum import (
     IPADDRESS_TYPE,
     NODE_STATUS,
-    NODE_STATUS_CHOICES,
+<<<<<<< TREE
+    NODE_STATUS_CHOICES,
+=======
+    NODE_STATUS_CHOICES,
+    NODE_STATUS_CHOICES_DICT,
+>>>>>>> MERGE-SOURCE
     )
 from maasserver.fields import (
     MAC,

=== modified file 'src/maasserver/api/tests/test_nodes.py'
=== modified file 'src/maasserver/bootresources.py'
=== modified file 'src/maasserver/forms.py'
=== modified file 'src/maasserver/forms_settings.py'
--- src/maasserver/forms_settings.py	2014-09-25 21:06:39 +0000
+++ src/maasserver/forms_settings.py	2014-11-06 01:24:10 +0000
@@ -266,6 +266,16 @@
                 "Erase nodes' disks prior to releasing.")
         }
     },
+    'enable_dhcp_discovery_on_unconfigured_interfaces': {
+        'default': False,
+        'form': forms.BooleanField,
+        'form_kwargs': {
+            'required': False,
+            'label': (
+                "Perform DHCP discovery on unconfigured network "
+                "interfaces of commissioning nodes."),
+        }
+    },
 }
 
 

=== modified file 'src/maasserver/middleware.py'
=== modified file 'src/maasserver/models/config.py'
--- src/maasserver/models/config.py	2014-10-08 20:41:31 +0000
+++ src/maasserver/models/config.py	2014-11-06 01:24:10 +0000
@@ -60,6 +60,7 @@
         # Third Party
         'enable_third_party_drivers': True,
         'enable_disk_erasing_on_release': False,
+        'enable_dhcp_discovery_on_unconfigured_interfaces': True,
         ## /settings
         }
 

=== modified file 'src/maasserver/models/node.py'
--- src/maasserver/models/node.py	2014-11-05 23:39:16 +0000
+++ src/maasserver/models/node.py	2014-11-06 01:24:10 +0000
@@ -319,6 +319,146 @@
         available_nodes = self.get_nodes(for_user, NODE_PERMISSION.VIEW)
         return available_nodes.filter(status=NODE_STATUS.READY)
 
+<<<<<<< TREE
+=======
+    def stop_nodes(self, ids, by_user, stop_mode='hard'):
+        """Request on given user's behalf that the given nodes be shut down.
+
+        Shutdown is only requested for nodes that the user has ownership
+        privileges for; any other nodes in the request are ignored.
+
+        :param ids: The `system_id` values for nodes to be shut down.
+        :type ids: Sequence
+        :param by_user: Requesting user.
+        :type by_user: User_
+        :param stop_mode: Power off mode - usually 'soft' or 'hard'.
+        :type stop_mode: unicode
+        :return: Those Nodes for which shutdown was actually requested.
+        :rtype: list
+        """
+        # Obtain node model objects for each node specified.
+        nodes = self.get_nodes(by_user, NODE_PERMISSION.EDIT, ids=ids)
+
+        # Helper function to whittle the list of nodes down to those that we
+        # can actually stop, and keep hold of their power control info.
+        def gen_power_info(nodes):
+            for node in nodes:
+                power_info = node.get_effective_power_info()
+                if power_info.can_be_stopped:
+                    # Smuggle in a hint about how to power-off the node.
+                    power_info.power_parameters['power_off_mode'] = stop_mode
+                    yield node, power_info
+
+        # Create info that we can pass into the reactor (no model objects).
+        nodes_stop_info = list(
+            (node.system_id, node.hostname, node.nodegroup.uuid, power_info)
+            for node, power_info in gen_power_info(nodes))
+        powered_systems = [
+            system_id for system_id, _, _, _ in nodes_stop_info]
+
+        # Request that these nodes be powered off and wait for the
+        # commands to return or fail.
+        deferreds = power_off_nodes(nodes_stop_info).viewvalues()
+        wait_for_power_commands(deferreds)
+
+        # Return a list of those nodes that we've sent power commands for.
+        return list(
+            node for node in nodes if node.system_id in powered_systems)
+
+    def start_nodes(self, ids, by_user, user_data=None):
+        """Request on given user's behalf that the given nodes be started up.
+
+        Power-on is only requested for nodes that the user has ownership
+        privileges for; any other nodes in the request are ignored.
+
+        Nodes are also ignored if they don't have a valid power type
+        configured.
+
+        :param ids: The `system_id` values for nodes to be started.
+        :type ids: Sequence
+        :param by_user: Requesting user.
+        :type by_user: User_
+        :param user_data: Optional blob of user-data to be made available to
+            the nodes through the metadata service.  If not given, any
+            previous user data is used.
+        :type user_data: unicode
+        :return: Those Nodes for which power-on was actually requested.
+        :rtype: list
+
+        :raises MultipleFailures: When there are failures originating from a
+            remote process. There could be one or more failures -- it's not
+            strictly *multiple* -- but they do all originate from comms with
+            remote processes.
+        :raises: `StaticIPAddressExhaustion` if there are not enough IP
+            addresses left in the static range..
+        """
+        # Avoid circular imports.
+        from metadataserver.models import NodeUserData
+
+        # Obtain node model objects for each node specified.
+        nodes = self.get_nodes(by_user, NODE_PERMISSION.EDIT, ids=ids)
+
+        # Record the same user data for all nodes we've been *requested* to
+        # start, regardless of whether or not we actually can; the user may
+        # choose to manually start them.
+        NodeUserData.objects.bulk_set_user_data(nodes, user_data)
+
+        # Claim static IP addresses for all nodes we've been *requested* to
+        # start, such that they're recorded in the database. This results in a
+        # mapping of nodegroups to (ips, macs).
+        static_mappings = defaultdict(dict)
+        for node in nodes:
+            if node.status == NODE_STATUS.ALLOCATED:
+                claims = node.claim_static_ip_addresses()
+                # If the PXE mac is on a managed interface then we can ask
+                # the cluster to generate the DHCP host map(s).
+                if node.is_pxe_mac_on_managed_interface():
+                    static_mappings[node.nodegroup].update(claims)
+                node.start_deployment()
+
+        # XXX 2014-06-17 bigjools bug=1330765
+        # If the above fails it needs to release the static IPs back to the
+        # pool. An enclosing transaction or savepoint from the caller may take
+        # care of this, given that a serious problem above will result in an
+        # exception. If we're being belt-n-braces though it ought to clear up
+        # before returning too. As part of the robustness work coming up, it
+        # also needs to inform the user.
+
+        # Update host maps and wait for them so that we can report failures
+        # directly to the caller.
+        update_host_maps_failures = list(update_host_maps(static_mappings))
+        if len(update_host_maps_failures) != 0:
+            raise MultipleFailures(*update_host_maps_failures)
+
+        # Update the DNS zone with the new static IP info as necessary.
+        from maasserver.dns.config import change_dns_zones
+        change_dns_zones({node.nodegroup for node in nodes})
+
+        # Helper function to whittle the list of nodes down to those that we
+        # can actually start, and keep hold of their power control info.
+        def gen_power_info(nodes):
+            for node in nodes:
+                power_info = node.get_effective_power_info()
+                if power_info.can_be_started:
+                    yield node, power_info
+
+        # Create info that we can pass into the reactor (no model objects).
+        nodes_start_info = list(
+            (node.system_id, node.hostname, node.nodegroup.uuid, power_info)
+            for node, power_info in gen_power_info(nodes))
+        powered_systems = [
+            system_id for system_id, _, _, _ in nodes_start_info]
+
+        # Request that these nodes be powered off and wait for the
+        # commands to return or fail.
+        deferreds = power_on_nodes(nodes_start_info).viewvalues()
+        wait_for_power_commands(deferreds)
+
+        # Return a list of those nodes that we've sent power commands for.
+        return list(
+            node for node in nodes if node.system_id in powered_systems)
+
+>>>>>>> MERGE-SOURCE
 
 def patch_pgarray_types():
     """Monkey-patch incompatibility with recent versions of `djorm_pgarray`.
@@ -1227,6 +1367,7 @@
         self.distro_series = ''
         self.license_key = ''
         self.save()
+<<<<<<< TREE
 
         # Clear installation results
         NodeResult.objects.filter(
@@ -1239,6 +1380,16 @@
         from maasserver.dns.config import change_dns_zones
         change_dns_zones([self.nodegroup])
 
+=======
+
+        # Do these after updating the node to avoid creating deadlocks with
+        # other node editing operations.
+        deallocated_ips = StaticIPAddress.objects.deallocate_by_node(self)
+        self.delete_host_maps(deallocated_ips)
+        from maasserver.dns.config import change_dns_zones
+        change_dns_zones([self.nodegroup])
+
+>>>>>>> MERGE-SOURCE
         # We explicitly commit here because during bulk node actions we
         # want to make sure that each successful state transition is
         # recorded in the DB.
@@ -1412,6 +1563,7 @@
             return self.pxe_mac
 
         return self.macaddress_set.first()
+<<<<<<< TREE
 
     def is_pxe_mac_on_managed_interface(self):
         pxe_mac = self.get_pxe_mac()
@@ -1541,3 +1693,13 @@
         d = power_off_nodes([stop_info]).values().pop()
         wait_for_power_command(d)
         return True
+=======
+
+    def is_pxe_mac_on_managed_interface(self):
+        pxe_mac = self.get_pxe_mac()
+        if pxe_mac is not None:
+            cluster_interface = pxe_mac.cluster_interface
+            if cluster_interface is not None:
+                return cluster_interface.is_managed
+        return False
+>>>>>>> MERGE-SOURCE

=== modified file 'src/maasserver/models/staticipaddress.py'
=== modified file 'src/maasserver/models/tests/test_bootsource.py'
--- src/maasserver/models/tests/test_bootsource.py	2014-11-03 01:36:42 +0000
+++ src/maasserver/models/tests/test_bootsource.py	2014-11-06 01:24:10 +0000
@@ -15,6 +15,7 @@
 __all__ = []
 
 import os
+from unittest import skip
 
 from django.core.exceptions import ValidationError
 from maasserver.bootsources import cache_boot_sources
@@ -98,6 +99,9 @@
             [],
             boot_source_dict['selections'])
 
+    # XXX: GavinPanella 2014-10-28 bug=1376317: This test is fragile, possibly
+    # due to isolation issues.
+    @skip("Possible isolation issues")
     def test_calls_cache_boot_sources_on_create(self):
         mock_callLater = self.patch(reactor, 'callLater')
         BootSource.objects.create(

=== modified file 'src/maasserver/models/tests/test_node.py'
--- src/maasserver/models/tests/test_node.py	2014-11-05 15:28:40 +0000
+++ src/maasserver/models/tests/test_node.py	2014-11-06 01:24:10 +0000
@@ -19,6 +19,7 @@
     timedelta,
     )
 import random
+from unittest import skip
 
 import crochet
 from django.core.exceptions import ValidationError
@@ -2083,6 +2084,422 @@
         self.assertThat(erase_mock, MockNotCalled())
 
 
+<<<<<<< TREE
+=======
+class NodeManagerTest_StartNodes(MAASServerTestCase):
+
+    def setUp(self):
+        super(NodeManagerTest_StartNodes, self).setUp()
+        self.useFixture(RegionEventLoopFixture("rpc"))
+        self.useFixture(RunningEventLoopFixture())
+        self.rpc_fixture = self.useFixture(MockLiveRegionToClusterRPCFixture())
+
+    def prepare_rpc_to_cluster(self, nodegroup):
+        protocol = self.rpc_fixture.makeCluster(
+            nodegroup, cluster_module.CreateHostMaps, cluster_module.PowerOn,
+            cluster_module.StartMonitors)
+        protocol.CreateHostMaps.side_effect = always_succeed_with({})
+        protocol.StartMonitors.side_effect = always_succeed_with({})
+        protocol.PowerOn.side_effect = always_succeed_with({})
+        return protocol
+
+    def make_acquired_nodes_with_macs(self, user, nodegroup=None, count=3):
+        nodes = []
+        for _ in xrange(count):
+            node = factory.make_node_with_mac_attached_to_nodegroupinterface(
+                nodegroup=nodegroup, status=NODE_STATUS.READY)
+            self.prepare_rpc_to_cluster(node.nodegroup)
+            node.acquire(user)
+            nodes.append(node)
+        return nodes
+
+    def test__sets_user_data(self):
+        user = factory.make_User()
+        nodegroup = factory.make_NodeGroup()
+        self.prepare_rpc_to_cluster(nodegroup)
+        nodes = self.make_acquired_nodes_with_macs(user, nodegroup)
+        user_data = factory.make_bytes()
+
+        with TwistedLoggerFixture() as twisted_log:
+            Node.objects.start_nodes(
+                list(node.system_id for node in nodes),
+                user, user_data=user_data)
+
+        # All three nodes have been given the same user data.
+        nuds = NodeUserData.objects.filter(
+            node_id__in=(node.id for node in nodes))
+        self.assertEqual({user_data}, {nud.data for nud in nuds})
+        # No complaints are made to the Twisted log.
+        self.assertFalse(twisted_log.containsError(), twisted_log.output)
+
+    def test__resets_user_data(self):
+        user = factory.make_User()
+        nodegroup = factory.make_NodeGroup()
+        self.prepare_rpc_to_cluster(nodegroup)
+        nodes = self.make_acquired_nodes_with_macs(user, nodegroup)
+
+        with TwistedLoggerFixture() as twisted_log:
+            Node.objects.start_nodes(
+                list(node.system_id for node in nodes),
+                user, user_data=None)
+
+        # All three nodes have been given the same user data.
+        nuds = NodeUserData.objects.filter(
+            node_id__in=(node.id for node in nodes))
+        self.assertThat(list(nuds), HasLength(0))
+        # No complaints are made to the Twisted log.
+        self.assertFalse(twisted_log.containsError(), twisted_log.output)
+
+    def test__claims_static_ip_addresses(self):
+        user = factory.make_User()
+        nodegroup = factory.make_NodeGroup()
+        self.prepare_rpc_to_cluster(nodegroup)
+        nodes = self.make_acquired_nodes_with_macs(user, nodegroup)
+
+        claim_static_ip_addresses = self.patch_autospec(
+            Node, "claim_static_ip_addresses", spec_set=False)
+        claim_static_ip_addresses.return_value = {}
+
+        with TwistedLoggerFixture() as twisted_log:
+            Node.objects.start_nodes(
+                list(node.system_id for node in nodes), user)
+
+        for node in nodes:
+            self.expectThat(claim_static_ip_addresses, MockAnyCall(node))
+        # No complaints are made to the Twisted log.
+        self.assertFalse(twisted_log.containsError(), twisted_log.output)
+
+    def test__claims_static_ip_addresses_for_allocated_nodes_only(self):
+        user = factory.make_User()
+        nodegroup = factory.make_NodeGroup()
+        self.prepare_rpc_to_cluster(nodegroup)
+        nodes = self.make_acquired_nodes_with_macs(user, nodegroup, count=2)
+
+        # Change the status of the first node to something other than
+        # allocated.
+        broken_node, allocated_node = nodes
+        broken_node.status = NODE_STATUS.BROKEN
+        broken_node.save()
+
+        claim_static_ip_addresses = self.patch_autospec(
+            Node, "claim_static_ip_addresses", spec_set=False)
+        claim_static_ip_addresses.return_value = {}
+
+        with TwistedLoggerFixture() as twisted_log:
+            Node.objects.start_nodes(
+                list(node.system_id for node in nodes), user)
+
+        # Only one call is made to claim_static_ip_addresses(), for the
+        # still-allocated node.
+        self.assertThat(
+            claim_static_ip_addresses,
+            MockCalledOnceWith(allocated_node))
+        # No complaints are made to the Twisted log.
+        self.assertFalse(twisted_log.containsError(), twisted_log.output)
+
+    def test__updates_host_maps(self):
+        user = factory.make_User()
+        nodes = self.make_acquired_nodes_with_macs(user)
+
+        update_host_maps = self.patch(node_module, "update_host_maps")
+        update_host_maps.return_value = []  # No failures.
+
+        with TwistedLoggerFixture() as twisted_log:
+            Node.objects.start_nodes(
+                list(node.system_id for node in nodes), user)
+
+        # Host maps are updated.
+        self.assertThat(
+            update_host_maps, MockCalledOnceWith({
+                node.nodegroup: {
+                    ip_address.ip: mac.mac_address
+                    for ip_address in mac.ip_addresses.all()
+                }
+                for node in nodes
+                for mac in node.mac_addresses_on_managed_interfaces()
+            }))
+        # No complaints are made to the Twisted log.
+        self.assertFalse(twisted_log.containsError(), twisted_log.output)
+
+    def test__propagates_errors_when_updating_host_maps(self):
+        user = factory.make_User()
+        nodes = self.make_acquired_nodes_with_macs(user)
+
+        update_host_maps = self.patch(node_module, "update_host_maps")
+        update_host_maps.return_value = [
+            Failure(AssertionError("That is so not true")),
+            Failure(ZeroDivisionError("I cannot defy mathematics")),
+        ]
+
+        with TwistedLoggerFixture() as twisted_log:
+            error = self.assertRaises(
+                MultipleFailures, Node.objects.start_nodes,
+                list(node.system_id for node in nodes), user)
+
+        self.assertSequenceEqual(
+            update_host_maps.return_value, error.args)
+
+        # No complaints are made to the Twisted log.
+        self.assertFalse(twisted_log.containsError(), twisted_log.output)
+
+    def test__updates_dns(self):
+        user = factory.make_User()
+        nodes = self.make_acquired_nodes_with_macs(user)
+
+        change_dns_zones = self.patch(dns_config, "change_dns_zones")
+
+        with TwistedLoggerFixture() as twisted_log:
+            Node.objects.start_nodes(
+                list(node.system_id for node in nodes), user)
+
+        self.assertThat(
+            change_dns_zones, MockCalledOnceWith(
+                {node.nodegroup for node in nodes}))
+
+        # No complaints are made to the Twisted log.
+        self.assertFalse(twisted_log.containsError(), twisted_log.output)
+
+    def test__starts_nodes(self):
+        user = factory.make_User()
+        nodes = self.make_acquired_nodes_with_macs(user)
+        power_infos = list(
+            node.get_effective_power_info()
+            for node in nodes)
+
+        power_on_nodes = self.patch(node_module, "power_on_nodes")
+        power_on_nodes.return_value = {}
+
+        with TwistedLoggerFixture() as twisted_log:
+            Node.objects.start_nodes(
+                list(node.system_id for node in nodes), user)
+
+        self.assertThat(power_on_nodes, MockCalledOnceWith(ANY))
+
+        nodes_start_info_observed = power_on_nodes.call_args[0][0]
+        nodes_start_info_expected = [
+            (node.system_id, node.hostname, node.nodegroup.uuid, power_info)
+            for node, power_info in izip(nodes, power_infos)
+        ]
+
+        # If the following fails the diff is big, but it's useful.
+        self.maxDiff = None
+
+        self.assertItemsEqual(
+            nodes_start_info_expected,
+            nodes_start_info_observed)
+
+        # No complaints are made to the Twisted log.
+        self.assertFalse(twisted_log.containsError(), twisted_log.output)
+
+    def test__raises_failures_for_nodes_that_cannot_be_started(self):
+        power_on_nodes = self.patch(node_module, "power_on_nodes")
+        power_on_nodes.return_value = {
+            factory.make_name("system_id"): defer.fail(
+                ZeroDivisionError("Defiance is futile")),
+            factory.make_name("system_id"): defer.succeed({}),
+        }
+
+        failures = self.assertRaises(
+            MultipleFailures, Node.objects.start_nodes, [],
+            factory.make_User())
+        [failure] = failures.args
+        self.assertThat(failure.value, IsInstance(ZeroDivisionError))
+
+    def test__marks_allocated_node_as_deploying(self):
+        user = factory.make_User()
+        [node] = self.make_acquired_nodes_with_macs(user, count=1)
+        nodes_started = Node.objects.start_nodes([node.system_id], user)
+        self.assertItemsEqual([node], nodes_started)
+        self.assertEqual(
+            NODE_STATUS.DEPLOYING, reload_object(node).status)
+
+    def test__does_not_change_state_of_deployed_node(self):
+        user = factory.make_User()
+        node = factory.make_Node(
+            power_type='ether_wake', status=NODE_STATUS.DEPLOYED,
+            owner=user)
+        factory.make_MACAddress(node=node)
+        power_on_nodes = self.patch(node_module, "power_on_nodes")
+        power_on_nodes.return_value = {
+            node.system_id: defer.succeed({}),
+        }
+        nodes_started = Node.objects.start_nodes([node.system_id], user)
+        self.assertItemsEqual([node], nodes_started)
+        self.assertEqual(
+            NODE_STATUS.DEPLOYED, reload_object(node).status)
+
+    def test__only_returns_nodes_for_which_power_commands_have_been_sent(self):
+        user = factory.make_User()
+        node1, node2 = self.make_acquired_nodes_with_macs(user, count=2)
+        node1.power_type = 'ether_wake'  # Can be started.
+        node1.save()
+        node2.power_type = ''  # Undefined power type, cannot be started.
+        node2.save()
+        nodes_started = Node.objects.start_nodes(
+            [node1.system_id, node2.system_id], user)
+        self.assertItemsEqual([node1], nodes_started)
+
+    def test__does_not_try_to_start_nodes_not_allocated_to_user(self):
+        user1 = factory.make_User()
+        [node1] = self.make_acquired_nodes_with_macs(user1, count=1)
+        node1.power_type = 'ether_wake'  # can be started.
+        node1.save()
+        user2 = factory.make_User()
+        [node2] = self.make_acquired_nodes_with_macs(user2, count=1)
+        node2.power_type = 'ether_wake'  # can be started.
+        node2.save()
+
+        self.patch(node_module, 'power_on_nodes')
+        self.patch(node_module, 'wait_for_power_commands')
+        nodes_started = Node.objects.start_nodes(
+            [node1.system_id, node2.system_id], user1)
+
+        # Since no power commands were sent to the node, it isn't
+        # returned by start_nodes().
+        # test__only_returns_nodes_for_which_power_commands_have_been_sent()
+        # demonstrates this behaviour.
+        self.assertItemsEqual([node1], nodes_started)
+
+    # XXX: GavinPanella 2014-10-30 bug=1387696: Flaky test, returning wrong IP
+    # addresses. Appears unrelated to the changes that highlighted this.
+    @skip("Flaky; returns incorrect IP addresses")
+    def test__does_not_generate_host_maps_if_not_on_managed_interface(self):
+        cluster = factory.make_NodeGroup()
+        managed_interface = factory.make_NodeGroupInterface(
+            nodegroup=cluster,
+            management=NODEGROUPINTERFACE_MANAGEMENT.DHCP_AND_DNS)
+        unmanaged_interface = factory.make_NodeGroupInterface(
+            nodegroup=cluster,
+            management=NODEGROUPINTERFACE_MANAGEMENT.DEFAULT)
+        user = factory.make_User()
+        [node1, node2] = self.make_acquired_nodes_with_macs(
+            user, nodegroup=cluster, count=2)
+        # Give the node a PXE MAC address on the cluster's interface.
+        node1_mac = node1.get_pxe_mac()
+        node1_mac.cluster_interface = managed_interface
+        node1_mac.save()
+        node2_mac = node1.get_pxe_mac()
+        node2_mac.cluster_interface = unmanaged_interface
+        node2_mac.save()
+
+        node1_ip = factory.make_ipv4_address()
+        claim_static_ip_addresses = self.patch(
+            node_module.Node, 'claim_static_ip_addresses')
+        claim_static_ip_addresses.side_effect = [
+            [(node1_ip, node1_mac.mac_address)],
+            [(factory.make_ipv4_address(), node2_mac.mac_address)],
+            ]
+
+        update_host_maps = self.patch(node_module, "update_host_maps")
+        Node.objects.start_nodes([node1.system_id, node2.system_id], user)
+        self.expectThat(update_host_maps, MockCalledOnceWith(ANY))
+
+        observed_static_mappings = update_host_maps.call_args[0][0]
+
+        [observed_cluster] = observed_static_mappings.keys()
+        self.expectThat(observed_cluster.uuid, Equals(cluster.uuid))
+
+        observed_claims = observed_static_mappings.values()
+        self.expectThat(
+            observed_claims,
+            Equals([{node1_ip: node1_mac.mac_address}]))
+
+
+class NodeManagerTest_StopNodes(MAASServerTestCase):
+
+    def make_nodes_with_macs(self, user, nodegroup=None, count=3):
+        nodes = []
+        for _ in xrange(count):
+            node = factory.make_node_with_mac_attached_to_nodegroupinterface(
+                nodegroup=nodegroup, status=NODE_STATUS.READY,
+                power_type='virsh')
+            node.acquire(user)
+            nodes.append(node)
+        return nodes
+
+    def test_stop_nodes_stops_nodes(self):
+        wait_for_power_commands = self.patch_autospec(
+            node_module, 'wait_for_power_commands')
+        power_off_nodes = self.patch_autospec(node_module, "power_off_nodes")
+        power_off_nodes.side_effect = lambda nodes: {
+            system_id: Deferred() for system_id, _, _, _ in nodes}
+
+        user = factory.make_User()
+        nodes = self.make_nodes_with_macs(user)
+        power_infos = list(node.get_effective_power_info() for node in nodes)
+
+        stop_mode = factory.make_name('stop-mode')
+        nodes_stopped = Node.objects.stop_nodes(
+            list(node.system_id for node in nodes), user, stop_mode)
+
+        self.assertItemsEqual(nodes, nodes_stopped)
+        self.assertThat(power_off_nodes, MockCalledOnceWith(ANY))
+        self.assertThat(wait_for_power_commands, MockCalledOnceWith(ANY))
+
+        nodes_stop_info_observed = power_off_nodes.call_args[0][0]
+        nodes_stop_info_expected = [
+            (node.system_id, node.hostname, node.nodegroup.uuid, power_info)
+            for node, power_info in izip(nodes, power_infos)
+        ]
+
+        # The stop mode is added into the power info that's passed.
+        for _, _, _, power_info in nodes_stop_info_expected:
+            power_info.power_parameters['power_off_mode'] = stop_mode
+
+        # If the following fails the diff is big, but it's useful.
+        self.maxDiff = None
+
+        self.assertItemsEqual(
+            nodes_stop_info_expected,
+            nodes_stop_info_observed)
+
+    def test_stop_nodes_ignores_uneditable_nodes(self):
+        owner = factory.make_User()
+        nodes = self.make_nodes_with_macs(owner)
+
+        user = factory.make_User()
+        nodes_stopped = Node.objects.stop_nodes(
+            list(node.system_id for node in nodes), user)
+
+        self.assertItemsEqual([], nodes_stopped)
+
+    def test_stop_nodes_does_not_attempt_power_off_if_no_power_type(self):
+        # If the node has a power_type set to UNKNOWN_POWER_TYPE, stop_nodes()
+        # won't attempt to power it off.
+        user = factory.make_User()
+        [node] = self.make_nodes_with_macs(user, count=1)
+        node.power_type = ""
+        node.save()
+
+        nodes_stopped = Node.objects.stop_nodes([node.system_id], user)
+        self.assertItemsEqual([], nodes_stopped)
+
+    def test_stop_nodes_does_not_attempt_power_off_if_cannot_be_stopped(self):
+        # If the node has a power_type that MAAS knows stopping does not work,
+        # stop_nodes() won't attempt to power it off.
+        user = factory.make_User()
+        [node] = self.make_nodes_with_macs(user, count=1)
+        node.power_type = "ether_wake"
+        node.save()
+
+        nodes_stopped = Node.objects.stop_nodes([node.system_id], user)
+        self.assertItemsEqual([], nodes_stopped)
+
+    def test__raises_failures_for_nodes_that_cannot_be_stopped(self):
+        power_off_nodes = self.patch(node_module, "power_off_nodes")
+        power_off_nodes.return_value = {
+            factory.make_name("system_id"): defer.fail(
+                ZeroDivisionError("Ee by gum lad, that's a rum 'un.")),
+            factory.make_name("system_id"): defer.succeed({}),
+        }
+
+        failures = self.assertRaises(
+            MultipleFailures, Node.objects.stop_nodes, [], factory.make_User())
+        [failure] = failures.args
+        self.assertThat(failure.value, IsInstance(ZeroDivisionError))
+
+
+>>>>>>> MERGE-SOURCE
 class TestNodeTransitionMonitors(MAASServerTestCase):
 
     def prepare_rpc(self):

=== modified file 'src/maasserver/node_action.py'
=== modified file 'src/maasserver/node_status.py'
=== modified file 'src/maasserver/tests/test_bootresources.py'
--- src/maasserver/tests/test_bootresources.py	2014-11-03 01:10:28 +0000
+++ src/maasserver/tests/test_bootresources.py	2014-11-06 01:24:10 +0000
@@ -1063,8 +1063,13 @@
             bootresources, 'cache_boot_sources')
         write_all_keyrings = self.patch(
             bootresources, 'write_all_keyrings')
+<<<<<<< TREE
         write_all_keyrings.return_value = [sentinel.source]
         image_descriptions = self.patch(
+=======
+        fake_write_all_keyrings.return_value = [sentinel.source]
+        fake_image_descriptions = self.patch(
+>>>>>>> MERGE-SOURCE
             bootresources, 'download_all_image_descriptions')
         descriptions = Mock()
         descriptions.is_empty.return_value = False
@@ -1082,15 +1087,29 @@
         self.expectThat(
             write_all_keyrings,
             MockCalledOnceWith(ANY, []))
+<<<<<<< TREE
         self.expectThat(
             image_descriptions,
             MockCalledOnceWith([sentinel.source]))
         self.expectThat(
             map_products,
+=======
+        self.assertThat(
+            fake_image_descriptions,
+            MockCalledOnceWith([sentinel.source]))
+        self.assertThat(
+            fake_map_products,
+>>>>>>> MERGE-SOURCE
             MockCalledOnceWith(descriptions))
+<<<<<<< TREE
         self.expectThat(
             download_all_boot_resources,
             MockCalledOnceWith([sentinel.source], sentinel.mapping))
+=======
+        self.assertThat(
+            fake_download_all_boot_resources,
+            MockCalledOnceWith([sentinel.source], sentinel.mapping))
+>>>>>>> MERGE-SOURCE
 
     def test__import_resources_has_env_GNUPGHOME_set(self):
         fake_image_descriptions = self.patch(

=== modified file 'src/maasserver/tests/test_middleware.py'
=== modified file 'src/maasserver/tests/test_node_action.py'
=== modified file 'src/maasserver/views/nodes.py'
--- src/maasserver/views/nodes.py	2014-11-05 15:24:51 +0000
+++ src/maasserver/views/nodes.py	2014-11-06 01:24:10 +0000
@@ -612,8 +612,13 @@
         context['nodecommissionresults'] = commissioning_results
 
         installation_results = NodeResult.objects.filter(
+<<<<<<< TREE
             node=node, result_type=RESULT_TYPE.INSTALLATION)
         if len(installation_results) > 1:
+=======
+            node=node, result_type=RESULT_TYPE.INSTALLING)
+        if len(installation_results) > 1:
+>>>>>>> MERGE-SOURCE
             for result in installation_results:
                 result.name = re.sub('[_.]', ' ', result.name)
             context['nodeinstallresults'] = installation_results

=== modified file 'src/maasserver/views/tests/test_nodes.py'
--- src/maasserver/views/tests/test_nodes.py	2014-11-05 15:24:51 +0000
+++ src/maasserver/views/tests/test_nodes.py	2014-11-06 01:24:10 +0000
@@ -1639,6 +1639,7 @@
         else:
             self.fail("Found more than one link: %s" % links)
 
+<<<<<<< TREE
     def get_installation_results_link(self, display):
         """Find the results link in `display`.
 
@@ -1656,6 +1657,25 @@
         elif len(links) > 1:
             return links
 
+=======
+    def get_installing_results_link(self, display):
+        """Find the results link in `display`.
+
+        :param display: Results display section for a node, as returned by
+            `request_results_display`.
+        :return: `lxml.html.HtmlElement` for the link to the node's
+            installation results, as found in `display`; or `None` if it was
+            not present.
+        """
+        links = display.cssselect('a')
+        if len(links) == 0:
+            return None
+        elif len(links) == 1:
+            return links[0]
+        elif len(links) > 1:
+            return links
+
+>>>>>>> MERGE-SOURCE
     def test_view_node_links_to_commissioning_results_if_appropriate(self):
         self.client_log_in(as_admin=True)
         result = factory.make_NodeResult_for_commissioning()
@@ -1735,8 +1755,13 @@
         self.logged_in_user = user
         result = factory.make_NodeResult_for_installation(node=node)
         section = self.request_results_display(
+<<<<<<< TREE
             result.node, RESULT_TYPE.INSTALLATION)
         link = self.get_installation_results_link(section)
+=======
+            result.node, RESULT_TYPE.INSTALLING)
+        link = self.get_installing_results_link(section)
+>>>>>>> MERGE-SOURCE
         self.assertNotIn(
             normalise_whitespace(link.text_content()),
             ('', None))
@@ -1778,6 +1803,35 @@
             ContainsAll(
                 [normalise_whitespace(link.text_content()) for link in links]))
 
+    def test_view_node_shows_single_installing_result(self):
+        self.client_log_in(as_admin=True)
+        result = factory.make_NodeResult_for_installing()
+        section = self.request_results_display(
+            result.node, RESULT_TYPE.INSTALLING)
+        link = self.get_installing_results_link(section)
+        self.assertEqual(
+            "install log",
+            normalise_whitespace(link.text_content()))
+
+    def test_view_node_shows_multiple_installing_results(self):
+        self.client_log_in(as_admin=True)
+        node = factory.make_Node()
+        num_results = randint(2, 5)
+        results_names = []
+        for _ in range(num_results):
+            node_result = factory.make_NodeResult_for_installing(node=node)
+            results_names.append(node_result.name)
+        section = self.request_results_display(
+            node, RESULT_TYPE.INSTALLING)
+        links = self.get_installing_results_link(section)
+        expected_results_names = list(reversed(results_names))
+        observed_results_names = list(
+            normalise_whitespace(link.text_content())
+            for link in links)
+        self.assertListEqual(
+            expected_results_names,
+            observed_results_names)
+
 
 class NodeListingSelectionJSControls(SeleniumTestCase):
 

=== modified file 'src/metadataserver/api.py'
--- src/metadataserver/api.py	2014-11-04 07:49:58 +0000
+++ src/metadataserver/api.py	2014-11-06 01:24:10 +0000
@@ -76,7 +76,7 @@
     NodeUserData,
     )
 from metadataserver.models.commissioningscript import (
-    BUILTIN_COMMISSIONING_SCRIPTS,
+    get_builtin_commissioning_scripts,
     )
 from piston.utils import rc
 from provisioningserver.events import (
@@ -229,8 +229,10 @@
         script_result = int(request.POST.get('script_result', 0))
         for name, uploaded_file in request.FILES.items():
             raw_content = uploaded_file.read()
-            if name in BUILTIN_COMMISSIONING_SCRIPTS:
-                postprocess_hook = BUILTIN_COMMISSIONING_SCRIPTS[name]['hook']
+            builtin_commissioning_scripts = (
+                get_builtin_commissioning_scripts())
+            if name in builtin_commissioning_scripts:
+                postprocess_hook = builtin_commissioning_scripts[name]['hook']
                 postprocess_hook(
                     node=node, output=raw_content,
                     exit_status=script_result)
@@ -280,8 +282,20 @@
 
         if node.status == NODE_STATUS.COMMISSIONING:
             self._store_commissioning_results(node, request)
+<<<<<<< TREE
             store_node_power_parameters(node, request)
             node.stop_transition_monitor()
+=======
+            # XXX 2014-10-21 newell, bug=1382075
+            # Auto detection for IPMI tries to save power parameters
+            # for Moonshot.  This causes issues if the node's power type
+            # is already MSCM as it uses SSH instead of IPMI.  This fix
+            # is temporary as power parameters should not be overwritten
+            # during commissioning because MAAS already has knowledge to
+            # boot the node.
+            if node.power_type != "mscm":
+                store_node_power_parameters(node, request)
+>>>>>>> MERGE-SOURCE
             target_status = self.signaling_statuses.get(status)
         elif node.status == NODE_STATUS.DEPLOYING:
             self._store_installation_results(node, request)

=== modified file 'src/metadataserver/models/commissioningscript.py'
--- src/metadataserver/models/commissioningscript.py	2014-10-29 22:37:02 +0000
+++ src/metadataserver/models/commissioningscript.py	2014-11-06 01:24:10 +0000
@@ -14,8 +14,8 @@
 
 __metaclass__ = type
 __all__ = [
-    'BUILTIN_COMMISSIONING_SCRIPTS',
     'CommissioningScript',
+    'get_builtin_commissioning_scripts',
     'inject_lldp_result',
     'inject_lshw_result',
     'inject_result',
@@ -43,6 +43,7 @@
     )
 from lxml import etree
 from maasserver.fields import MAC
+from maasserver.models import Config
 from maasserver.models.tag import Tag
 from metadataserver import DefaultMeta
 from metadataserver.enum import RESULT_TYPE
@@ -352,6 +353,7 @@
 LIST_MODALIASES_OUTPUT_NAME = '00-maas-04-list-modaliases.out'
 LIST_MODALIASES_SCRIPT = \
     'find /sys -name modalias -print0 | xargs -0 cat | sort -u'
+DHCP_UNCONFIGURED_INTERFACES_NAME = '00-maas-05-dhcp-unconfigured-ifaces'
 
 
 def null_hook(node, output, exit_status):
@@ -397,7 +399,7 @@
         'content': LIST_MODALIASES_SCRIPT.encode('ascii'),
         'hook': null_hook,
     },
-    '00-maas-05-dhcp-unconfigured-ifaces': {
+    DHCP_UNCONFIGURED_INTERFACES_NAME: {
         'content': make_function_call_script(dhcp_explore),
         'hook': null_hook,
     },
@@ -430,6 +432,20 @@
 add_names_to_scripts(BUILTIN_COMMISSIONING_SCRIPTS)
 
 
+def get_builtin_commissioning_scripts():
+    """Get the builtin commissioning scripts.
+
+    The builtin scripts exposed may vary based on config settings.
+    """
+    scripts = BUILTIN_COMMISSIONING_SCRIPTS.copy()
+
+    config_key = 'enable_dhcp_discovery_on_unconfigured_interfaces'
+    if not Config.objects.get_config(config_key):
+        del scripts[DHCP_UNCONFIGURED_INTERFACES_NAME]
+
+    return scripts
+
+
 def add_script_to_archive(tarball, name, content, mtime):
     """Add a commissioning script to an archive of commissioning scripts."""
     assert isinstance(content, bytes), "Script content must be binary."
@@ -447,7 +463,7 @@
     """Utility for the collection of `CommissioningScript`s."""
 
     def _iter_builtin_scripts(self):
-        for script in BUILTIN_COMMISSIONING_SCRIPTS.itervalues():
+        for script in get_builtin_commissioning_scripts().itervalues():
             yield script['name'], script['content']
 
     def _iter_user_scripts(self):
@@ -500,8 +516,9 @@
     NodeResult.objects.store_data(
         node, name, script_result=exit_status,
         result_type=RESULT_TYPE.COMMISSIONING, data=Bin(output))
-    if name in BUILTIN_COMMISSIONING_SCRIPTS:
-        postprocess_hook = BUILTIN_COMMISSIONING_SCRIPTS[name]['hook']
+    builtin_commissioning_scripts = get_builtin_commissioning_scripts()
+    if name in builtin_commissioning_scripts:
+        postprocess_hook = builtin_commissioning_scripts[name]['hook']
         postprocess_hook(node=node, output=output, exit_status=exit_status)
 
 

=== modified file 'src/metadataserver/models/tests/test_noderesults.py'
--- src/metadataserver/models/tests/test_noderesults.py	2014-10-29 22:37:02 +0000
+++ src/metadataserver/models/tests/test_noderesults.py	2014-11-06 01:24:10 +0000
@@ -35,6 +35,7 @@
 
 from fixtures import FakeLogger
 from maasserver.fields import MAC
+from maasserver.models import Config
 from maasserver.models.tag import Tag
 from maasserver.testing.factory import factory
 from maasserver.testing.orm import reload_object
@@ -55,7 +56,10 @@
     )
 from metadataserver.models.commissioningscript import (
     ARCHIVE_PREFIX,
+    BUILTIN_COMMISSIONING_SCRIPTS,
+    DHCP_UNCONFIGURED_INTERFACES_NAME,
     extract_router_mac_addresses,
+    get_builtin_commissioning_scripts,
     inject_lldp_result,
     inject_lshw_result,
     inject_result,
@@ -676,3 +680,26 @@
         logger = self.useFixture(FakeLogger(name='commissioningscript'))
         update_hardware_details(factory.make_Node(), b"garbage", exit_status=1)
         self.assertEqual("", logger.output)
+
+
+class TestGetBuiltinCommissioningScripts(MAASServerTestCase):
+
+    def test__includes_all_builtin_commissioning_scripts_by_default(self):
+        self.assertItemsEqual(
+            BUILTIN_COMMISSIONING_SCRIPTS,
+            get_builtin_commissioning_scripts(),
+        )
+
+    def test__excludes_dhcp_discovery_when_disabled(self):
+        Config.objects.set_config(
+            'enable_dhcp_discovery_on_unconfigured_interfaces', False)
+        self.assertNotIn(
+            DHCP_UNCONFIGURED_INTERFACES_NAME,
+            get_builtin_commissioning_scripts())
+
+    def test__includes_dhcp_discovery_when_enabled(self):
+        Config.objects.set_config(
+            'enable_dhcp_discovery_on_unconfigured_interfaces', True)
+        self.assertIn(
+            DHCP_UNCONFIGURED_INTERFACES_NAME,
+            get_builtin_commissioning_scripts())

=== modified file 'src/metadataserver/tests/test_api.py'
--- src/metadataserver/tests/test_api.py	2014-11-04 07:49:41 +0000
+++ src/metadataserver/tests/test_api.py	2014-11-06 01:24:10 +0000
@@ -812,6 +812,21 @@
         node = reload_object(node)
         self.assertEqual(0, len(node.tags.all()))
 
+    def test_signal_current_power_type_mscm_does_not_store_params(self):
+        node = factory.make_Node(
+            power_type="mscm", status=NODE_STATUS.COMMISSIONING)
+        client = make_node_client(node=node)
+        params = dict(
+            power_address=factory.make_string(),
+            power_user=factory.make_string(),
+            power_pass=factory.make_string())
+        response = call_signal(
+            client, power_type="moonshot", power_parameters=json.dumps(params))
+        self.assertEqual(httplib.OK, response.status_code, response.content)
+        node = reload_object(node)
+        self.assertEqual("mscm", node.power_type)
+        self.assertNotEqual(params, node.power_parameters)
+
     def test_signal_refuses_bad_power_type(self):
         node = factory.make_Node(status=NODE_STATUS.COMMISSIONING)
         client = make_node_client(node=node)

=== modified file 'src/provisioningserver/rpc/boot_images.py'
--- src/provisioningserver/rpc/boot_images.py	2014-10-30 13:17:51 +0000
+++ src/provisioningserver/rpc/boot_images.py	2014-11-06 01:24:10 +0000
@@ -18,8 +18,12 @@
     "is_import_boot_images_running",
     ]
 
+<<<<<<< TREE
 import os
 from urlparse import urlparse
+=======
+from urlparse import urlparse
+>>>>>>> MERGE-SOURCE
 
 from provisioningserver import concurrency
 from provisioningserver.auth import get_maas_user_gpghome
@@ -47,6 +51,16 @@
     return hosts
 
 
+def get_hosts_from_sources(sources):
+    """Return set of hosts that are contained in the given sources."""
+    hosts = set()
+    for source in sources:
+        url = urlparse(source['url'])
+        if url.hostname is not None:
+            hosts.add(url.hostname)
+    return hosts
+
+
 @synchronous
 def _run_import(sources, http_proxy=None, https_proxy=None):
     """Run the import.

=== modified file 'src/provisioningserver/rpc/clusterservice.py'
=== modified file 'src/provisioningserver/rpc/power.py'
=== modified file 'src/provisioningserver/rpc/tests/test_boot_images.py'
--- src/provisioningserver/rpc/tests/test_boot_images.py	2014-10-30 13:17:51 +0000
+++ src/provisioningserver/rpc/tests/test_boot_images.py	2014-11-06 01:24:10 +0000
@@ -76,6 +76,13 @@
         self.assertItemsEqual(hosts, get_hosts_from_sources(sources))
 
 
+class TestGetHostsFromSources(PservTestCase):
+
+    def test__returns_set_of_hosts_from_sources(self):
+        sources, hosts = make_sources()
+        self.assertItemsEqual(hosts, get_hosts_from_sources(sources))
+
+
 class TestRunImport(PservTestCase):
 
     def make_archive_url(self, name=None):

=== modified file 'src/provisioningserver/rpc/tests/test_clusterservice.py'
=== modified file 'src/provisioningserver/rpc/tests/test_power.py'