#opendaylight-meeting: kernel projects
Meeting started by rgoulding at 17:02:08 UTC
(full logs).
Meeting summary
- agenda bashing (rgoulding, 17:02:18)
- Infra issues (rgoulding,
17:02:36)
- re-enable archetype self testing (rgoulding,
17:05:27)
- infra issues (rgoulding, 17:05:39)
- three updates (rgoulding,
17:06:17)
- 1) last night there was severe disk i/o and
network slowdowns (rgoulding,
17:06:32)
- vexhost trying to give new hardware, and they
did this last night (rgoulding,
17:06:41)
- when they did this, the upgrade slowed
everything down due to sync costs (rgoulding,
17:07:01)
- hopefully will not see the very slow transfers
that were happening last night (rgoulding,
17:07:18)
- 2) filesystem ran out of disk inodes
(rgoulding,
17:07:25)
- the ext4 partition ran out of inodes for
nexus (rgoulding,
17:07:39)
- ran out of the ability to index the disk
(rgoulding,
17:07:54)
- xfs partitioned disk has been created
(rgoulding,
17:08:04)
- which can support this (rgoulding,
17:08:08)
- the sync will take a very long time to
sync (rgoulding,
17:08:15)
- some things have been deleted to helpf
accelerate the transition (rgoulding,
17:08:26)
- includes a lot of logs that are tiny, which
take up a ton of inodes (rgoulding,
17:08:39)
- should hold us over for the week or so until we
have the XFS system in place (rgoulding,
17:08:50)
- 3) issues with maven merge jobs (rgoulding,
17:09:01)
- globaljjb maven merge jobs is different than
what we used in ODL originally (freestyle) (rgoulding,
17:09:26)
- CLI tool is producing invalid timestamp for
files it is producing locally (rgoulding,
17:09:53)
- a bug has been discovered in maven 3.5.0 which
is similar to this issue, even though we aren’t using it (we use
3.3.9) (rgoulding,
17:10:15)
- reverted to ODL - specific merge jobs
(rgoulding,
17:10:23)
- these jobs were around for a month, which is
why it is confusing that we suddenly see this issue (rgoulding,
17:10:41)
- rovarga has a theory that it was probably due
to the fact that we had a limited number of patches actually getting
merged during the odlparent/yangtools migration (rgoulding,
17:11:31)
- now we have a huge backlog of patches, so we
are more likely to run into this issue (rgoulding,
17:11:53)
- in the meantime we can possibly detect and
correct the maven metadata (rgoulding,
17:13:31)
- we are going to wait on switching this until
the release is out the door (rgoulding,
17:13:59)
- if merge jobs continue to fail, please contact
zxiiro and he will look into it (rgoulding,
17:14:29)
- vorburger brings up a 4th possible error;
there are memory issues with validate-autorelease (rgoulding,
17:14:51)
- the VM template was changed, which could cause
this (rgoulding,
17:16:05)
- reverting validate-autorelease to use 2 core /
8 gig RAM templates will result in slower jobs, but they won’t crash
due to OOM at least (rgoulding,
17:21:41)
- https://logs.opendaylight.org/releng/vex-yul-odl-jenkins-1/mdsal-validate-autorelease-oxygen/181/console.log.gz
(rovarga_,
17:23:16)
- the above link was provided to show cpu/memory
analysis for validate-autorelease process (rgoulding,
17:24:13)
- it would appear that validate-autorelease is
very memory sensitive, but less CPU sensitive (rgoulding,
17:24:29)
- zxiiro will revert the template for
validate-autorelease, but not touch the others since they don’t
require the same resources (rgoulding,
17:24:55)
- further analysis of this and broader messaging
could help in the future (rgoulding,
17:26:36)
- archetype self-testing (rgoulding, 17:28:12)
- https://git.opendaylight.org/gerrit/#/c/67833/
(rgoulding,
17:28:15)
- 14:49:16
networkx.exception.NetworkXUnfeasible: Graph contains a cycle or
graph changed during iteration (rgoulding,
17:31:08)
- it appears there is a cycle (rgoulding,
17:31:20)
- https://github.com/opendaylight/controller/blob/master/opendaylight/archetypes/opendaylight-startup/src/main/resources/archetype-resources/features/odl-__artifactId__-rest/pom.xml#L35-L38
(rgoulding,
17:32:35)
- this causes the cyclic dependency loop
(rgoulding,
17:32:42)
- so this can’t actually work (with self-testing)
in controller (rgoulding,
17:34:20)
- since then controller depends on netconf, which
is not correct (rgoulding,
17:34:31)
- this self-test helps detect errors before
releasing (rgoulding,
17:35:31)
- and it is used a bunch more than what we had
imagined (rgoulding,
17:35:41)
- coretutorials may be a good landing spot for
this, since it is outside of autorelease (rgoulding,
17:36:35)
- i.e., this is how you generate the code
etc. (rgoulding,
17:36:50)
- coretutorials is currently broken (behind
nitrogen and oxygen currently) but Robert has some patches to fix
this (rgoulding,
17:37:14)
- tpantelis is a committer / PTL for it
(rgoulding,
17:37:58)
- rovarga is suggesting to move the archetype to
coretutorials (rgoulding,
17:39:27)
- with the tests and everything (rgoulding,
17:39:32)
- you can then evolve the archetype outside the
normal release process since you can release the artifacts outside
of autorelease (rgoulding,
17:39:51)
- would it make sense to keep in autorelease?
that would allow us to see it breaking live (rgoulding,
17:40:26)
- skitt brings up the fact that we could do it in
other ways (rgoulding,
17:41:06)
- https://git.opendaylight.org/gerrit/#/q/status:open+project:coretutorials+branch:stable/nitrogen
(rovarga_,
17:41:46)
- if it lives in autorelease, you cannot provide
bugfixes etc. outside of autorelease (rgoulding,
17:44:17)
- there are benefits and drawbacks associated
with all of this (rgoulding,
17:44:39)
- ACTION: vorburger to
move archetype to coretutorials (rgoulding,
17:45:54)
- propose a separate project for just the
archetype (rgoulding,
17:47:48)
- why don’t we just delete it? (rgoulding,
17:48:24)
- https://wiki.opendaylight.org/view/Project_Proposals:Main
(rovarga_,
17:51:47)
- wiki page describing what needs to do to
propose a new project (rovarga_,
17:52:04)
- this page will likely need to be reverted as
someone screwed the template (rgoulding,
17:56:21)
- tpantelis to look at
https://git.opendaylight.org/gerrit/#/c/67662/ (rgoulding,
18:00:03)
Meeting ended at 18:02:17 UTC
(full logs).
Action items
- vorburger to move archetype to coretutorials
Action items, by person
- vorburger
- vorburger to move archetype to coretutorials
People present (lines said)
- rgoulding (64)
- rovarga_ (4)
- odl_meetbot (3)
- vorburger (1)
Generated by MeetBot 0.1.4.