lp:~powersj/curtin/fix-vmtest-log-exception
- Get this branch:
- bzr branch lp:~powersj/curtin/fix-vmtest-log-exception
Branch merges
- Ryan Harper (community): Approve
- Server Team CI bot: Approve (continuous-integration)
-
Diff: 17 lines (+5/-2)1 file modifiedtests/vmtests/__init__.py (+5/-2)
Related bugs
Bug #1714592: vmtest: uncaught exception results in lage image files | Medium | Fix Released |
Related blueprints
Branch information
Recent revisions
- 520. By Ryan Harper
-
Update network cloud-init related skiptest dates, SRU still pending
Move skipdate to 2017-08-20 to allow more time for cloud-init SRU to land.
- 519. By Chad Smith
-
tests: Add CiTestCase common parent for all curtin tests.
This branch makes sure all unittests subclass from CiTestCase which
provides all unit tests with a tmp_dir and tmp_path creation. Drop
unittest-specific tempfile usage in favor of self.tmp_dir and
self.tmp_path.Adding this parent class gives us a single place to add test methods which
have general application across unit tests, like log tracking and tmp
directory cleanup. - 518. By Ryan Harper
-
vmtests: Remove force flag for centos curthooks
During development, and before our source images had updated
curtin hooks we needed this flag to push test installations down
the new code path.We've left it present for future development needs.
- 517. By Scott Moser
-
tools/jenkins-
runner: improve tgtd cleanup logic The previous tgtd cleanup logic was hard to follow at best. It was buggy
in that it would always kill the TGT_PID process even if it didn't start it.
both cleanup lines ended with:
<condition> || [ -n "${TGT_PID}" ] && kill -9 ${TGT_PID}
which will always run the 'kill' if TGT_PID is non-empty.Also:
* clean up the output of tools/jenkins-runner by redirecting
the find-tgt.d to a file so you don't see its debug messages.
* use modulus to get a port in the right range.
* improve the picking of a port and then checking to see if it was used.
I believe a race condition where tgt wasn't listening on the port when
netstat would run lead to false positives for failures where a
sleep would have sufficed.
* remove the tgt socket file(s) if it was created. - 516. By Chad Smith
-
tests: Drop EOL Wily Vivid and Yakkety tests.
We no longer carry test classes for end-of-
life/unsupporte d series. - 515. By Ryan Harper
-
Disable yum plugins when installing packages, update ca-certs for https
The default yum configured plugins includes the 'fastmirrors' plugin which
produces inconsistent results resulting in broken installs. Update the
ca-certificates package to prevent https mirror download failures. Add
some retries to the yum install commands to handle transient failures. - 514. By Scott Moser
-
Rename centos_
network_ curthooks -> centos_ apply_network_ config. Rename centos_
network_ curthooks and its upper-cased feature
CENTOS_NETWORK_ CURTHOOKS to centos_ apply_network_ config and
CENTOS_APPLY_NETWORK_ CONFIG respectively. Also drop the handling of 'cloudconfig' entries from that function.
Handling of cloudconfig is up to the curthooks, and that is being
added to maas-images built-in curthooks.The name centos_
apply_network_ config is left in curtin. cmd.curthooks.
We add a unittest to ensures that entry point is never lost.Also add unit test to ensure 'write_files' is available from futil.
- 513. By Scott Moser
-
tests: in centos_defaults use write_files for grub serial.
This changes centos defaults to pass a file into the target
with write_files and then use that program in late_commands to
modify grub1 and grub2 config files so that we get console
output to the serial console.Instead of running grub2-mkconfig it just updates both
the defaults file and the rendered /boot/grub2/grub.cfg. - 512. By Scott Moser
-
write_files: write files after extract, change write_files signature.
Instead of relying on curthooks to write files, write them during 'extract'.
The reason for this is that curthooks can be over-ridden by the image
that is being installed, but writing files is generic.The motivation for this was that I found centos images did not write files.
Also present here is a cleanup of 'write_files' signature and a move
from curtin.commands. curthooks to futil. It previously took a dictionary,
and operated on only the 'write_files' entry inside it. Now we take that
dictionary directly and change callers. The kwarg name 'target' was
changed to 'base_dir' to more correctly identify what this is.
Branch metadata
- Branch format:
- Branch format 7
- Repository format:
- Bazaar repository format 2a (needs bzr 1.16 or later)
- Stacked on:
- lp:~curtin-dev/curtin/trunk