nfs-charm

nfs-charm:master

  1. Git
  2. lp:nfs-charm
  3. master
Last commit made on 2019-07-15
Get this branch:
git clone -b master https://git.launchpad.net/nfs-charm

Branch merges

Branch information

Name:
master
Repository:
lp:nfs-charm

Recent commits

32cc982... by Mike Wilson <email address hidden>

Merge pull request #8 from cjwatson/fix-owncloud-example

Fix typo in example

414ad6e... by Colin Watson

Fix typo in example

88853da... by Mike Wilson <email address hidden>

Add active_units configuration (#7)

* Add active_units configuration

Assuming that the requiring charms support it, this allows reasonably
graceful migration from one NFS unit to another, along the lines of the
following procedure:

  $ juju config nfs active_units=<old unit ID>
  $ juju add-unit nfs
  $ juju config nfs active_units=none
  # wait for clients to unmount, then move storage to new unit
  $ juju config nfs active_units=<new unit ID>
  # wait for clients to mount
  $ juju remove-unit <old unit ID>
  $ juju config nfs active_units=

This involves moving publishing data on the mount relation to
`read_nfs_mounts`, which seems more correct anyway (we shouldn't publish
a mount until it has been exported).

* Add storage migration documentation

* Tweak how read_nfs_mounts gets the mount endpoint

The `endpoint.nfs.joined` flag isn't necessarily set when
`read_nfs_mounts` is called, so use endpoint_from_name('nfs') instead.

* Publish active unit from all units

In order that the mount interface and requiring charms don't need to
change to support `active_units` migration, we now use a peer relation
to find the address of the first active unit and publish that from all
units.

* Note abruptness of migration

7cd9c1c... by Colin Watson

Accept whitespace in active_units

f6ebb01... by Colin Watson

Simplify config handling in read_nfs_mounts

45b4064... by Colin Watson

Simplify read_nfs_mounts prologue

d7c0bb5... by Colin Watson

Stop breaking the mount interface's encapsulation

db6c2bb... by Colin Watson

Note abruptness of migration

be14b0d... by Colin Watson

Publish active unit from all units

In order that the mount interface and requiring charms don't need to
change to support `active_units` migration, we now use a peer relation
to find the address of the first active unit and publish that from all
units.

2bcd819... by Colin Watson

Tweak how read_nfs_mounts gets the mount endpoint

The `endpoint.nfs.joined` flag isn't necessarily set when
`read_nfs_mounts` is called, so use endpoint_from_name('nfs') instead.