#opendaylight-meeting: kernel projects
Meeting started by rgoulding at 16:01:52 UTC
(full logs).
Meeting summary
- agenda bashing (rgoulding, 16:02:01)
- https://bugs.opendaylight.org/show_bug.cgi?id=8839
(rgoulding,
16:02:13)
- JavaDoc discussion on odlparent-dev
(rgoulding,
16:02:18)
- if time & interest:
infrautils.MoreFutures.logFailure:
https://git.opendaylight.org/gerrit/#/c/60207/ (rgoulding,
16:02:26)
- https://git.opendaylight.org/gerrit/#/c/60207/
(rgoulding,
16:02:34)
- if time & interest:
infrautils.LogCaptureRule:
https://git.opendaylight.org/gerrit/#/c/60204/ (rgoulding,
16:02:41)
- https://git.opendaylight.org/gerrit/#/c/60204/
(rgoulding,
16:02:47)
- Bug 8839 (rgoulding, 16:03:59)
- https://bugs.opendaylight.org/show_bug.cgi?id=8839
(rgoulding,
16:04:11)
- ACTION: Sai and Luis
to let Atul Gosain know the NETCONF password encryption patch is
being reverted to unblock the Carbon-SR1 release (rgoulding,
16:06:22)
- perhaps before we merge these sorts of patches
we should rquire some sort of corresponding CSIT test (rgoulding,
16:07:24)
- ACTION: to test the
encryption mechanism in master (rgoulding,
16:07:37)
- ACTION: tcere to +2
netconf encryption patch andemail releng team to merge the
patch (rgoulding,
16:13:59)
- https://docs.google.com/spreadsheets/d/1VcB12FBiFV4GAEHZSspHBNxKI_9XugJp-6Qbbw20Omk/edit#gid=921315511
Carbon SR1 Tracking spreadsheet (zxiiro,
16:14:00)
- carbon tracking (rgoulding, 16:14:16)
- Thanh is concerned about 8805 since he is
unsure if thjey will require respin (rgoulding,
16:14:59)
- https://bugs.opendaylight.org/show_bug.cgi?id=8805 is
not a blocker (rgoulding,
16:15:39)
- https://bugs.opendaylight.org/show_bug.cgi?id=8805
(rgoulding,
16:15:41)
- JavaDoc discussion on odlparent-dev (rgoulding, 16:16:15)
- https://lists.opendaylight.org/pipermail/odlparent-dev/2017-July/001215.html
(rgoulding,
16:17:17)
- maven-javadoc-plugin and whether it is possible
for us to deploy to some site? what mechanism would we use to
deploy? (rgoulding,
16:17:43)
- zxiiro is unsure at the moment, but he
recommends pushes the artifacts to nexus so we can deploy when we
figure it out (rgoulding,
16:18:11)
- zxiiro if you have a deploy-site.xml, it will
trigger jenkins to build the javadoc site (rgoulding,
16:18:28)
- https://nexus.opendaylight.org/content/sites/site/
Maven Site (zxiiro,
16:18:39)
- rovarga this is where the aggregate javadocs
should be pushed out (rgoulding,
16:19:24)
- rovarga this runs aggregates recursively
(rgoulding,
16:19:36)
- https://nexus.opendaylight.org/content/sites/site/org.opendaylight.odlparent/boron/maven-site.html
current instructions for creating maven sites (zxiiro,
16:20:26)
- rovarga should we publish the javadocs by
release train? (rgoulding,
16:23:53)
- then we can reverse proxy to the appropriate
sites deployed on nexus (rgoulding,
16:26:38)
- zxiiro if we do it by version they only get
released when we release a new version of ODL (rgoulding,
16:27:01)
- zxiiro if we update now it gets pushed out
immediately since they are SNAPSHOT builds (rgoulding,
16:27:20)
- how much value is this worth? (rgoulding,
16:27:40)
- vorburger why do this every six months if we
can do it every change (rgoulding,
16:27:56)
- bug fixes also include javadoc
clarifications (rgoulding,
16:28:04)
- rovarga consistency of javadocs could become a
problem if we do it per merge on SNAPSHOT builds (rgoulding,
16:28:24)
- vorburger on the branch it will still be the
right versions, right? (rgoulding,
16:29:03)
- rovarga in cross project related javadoc
changes you may have stale links (rgoulding,
16:29:53)
- vorburger is this a corner case? (rgoulding,
16:30:00)
- rovarga doesn’t have to be code related.
javadocs have links that can break but not at compile time
(rgoulding,
16:30:24)
- this is related to the “@link” style
annotations (rgoulding,
16:30:32)
- vorburger says he would prefer having a site
with broken links rather than not having the site at all
(rgoulding,
16:31:03)
- zxirro we can rewrite maven sites at any
time (rgoulding,
16:31:17)
- can we wipe the crud that gets accumulated over
the release frame? (rgoulding,
16:32:54)
- it is possible (rgoulding,
16:33:03)
- job could be updated to do it
automatically (rgoulding,
16:33:08)
- rovarga says the site replacement ought to be
atomic (rgoulding,
16:33:58)
- zxiiro nexus only provides deploy mechanism,
not move mechanism (rgoulding,
16:34:19)
- ACTION: vorburger to
open a bug against releng/builder for wiping sites (rgoulding,
16:35:45)
- Karaf 4 Update (rgoulding, 16:39:51)
- ecelgp thinks that it is slow how projects are
coming into integration/distribution (rgoulding,
16:40:39)
- but it is happening still (rgoulding,
16:40:45)
- two questions 1) karaf4 issue with NETCONF and
dependent project (rgoulding,
16:41:43)
- 2) Bug 8824 - NETCONF request hangs when
rpc-rply has invalid xml (rgoulding,
16:42:33)
- https://bugs.opendaylight.org/show_bug.cgi?id=8803
(rgoulding,
16:43:26)
- rovarga need karaf log for this (rgoulding,
16:44:53)
- rovarga if it truly is a deadlock/busy wait
then we need a thread dump (rgoulding,
16:45:04)
- rovarga until it is further analyzed we
shouldn’t move it from openflowplugin to netconf because we don’t
have enough information to reproduce or analyze (rgoulding,
16:45:31)
- still one issue with config subsystem features
added to integration/distribution (rgoulding,
16:46:02)
- projects that sitll have dependency on NETCONF
need to hold off from joining integration/distribution unitl it
resolves this issue (rgoulding,
16:46:20)
- rovarga asks ecelgp if it affects nitrogen
only (rgoulding,
16:46:55)
- ecelgp it affects nitrogen only (rgoulding,
16:47:08)
- ACTION: ecelgp to
push karaf.log or thread dump depending on whetherCPU spikes to
100% (rgoulding,
16:47:26)
- if time & interest: infrautils.MoreFutures.logFailure: https://git.opendaylight.org/gerrit/#/c/60207/ (rgoulding, 16:47:34)
- https://git.opendaylight.org/gerrit/#/c/60207/
(rgoulding,
16:47:40)
- vorburger pushed this patch for netvirt, but
wants to make sure it is the right way and that there is nto a
simpler way to log if anything went wrong when doing
Transaction.submit() and ignore the returned future (rgoulding,
16:48:25)
- rovarga why are you ignoring the Future? That
would mean you have no recovery scenarios (rgoulding,
16:49:31)
- vorburger claims that most of the DS
interactions don’t have recovery scenarios (rgoulding,
16:49:56)
- vorburger a warning in the log is a lot better
than not getting anything at all (rgoulding,
16:50:11)
- rovarga would be much happier with people
explicitly adding these (rgoulding,
16:50:30)
- if you are doing high performance stuff you
probably don’t want to be using this thing and adding your own
callbacks on top of it? (rgoulding,
16:50:59)
- if you are adding in either recovery or succes
on top of this logging, then it will degrade performance
(rgoulding,
16:51:29)
- vorburger javadoc should contain a big fat bold
statement saying that you should be doing this on your own isntead
of using this (rgoulding,
16:51:49)
- skitt instead of converting all the current
code to use this new wrapper that adds a logging handler, make that
explicit in the code everywhere that needs to have error
handling/recovery (rgoulding,
16:52:47)
- vorburger that assumes people know how to do
recovery code (rgoulding,
16:53:10)
- vorburger, they haven’t been in 3-4
years (rgoulding,
16:53:19)
- skitt says this will hide the real problem and
no one will actually know what they are supposed to do (rgoulding,
16:53:40)
- rovarga the long term strategy should probably
involve having an annotation to submit and ensure the return value
is not ignored (rgoulding,
16:54:23)
- this will at least trigger an eclipse
warning (rgoulding,
16:54:38)
- skitt this may have people put related
operations into a single transactionchain instead of many small
transactions (rgoulding,
16:55:07)
- vorburger thinks this may be a bit
optimistic (rgoulding,
16:55:29)
- rovarga this is really useful for single
transactions, but less useful for TransactionChain(s) (rgoulding,
16:55:52)
- rovarga you can hook an error listener to a
TransactionChain for error recovery (rgoulding,
16:56:08)
- vorburger many thousands of line of code does
not use TransactionChain(s) (rgoulding,
16:56:20)
- skitt even if there is deprecation warning or
javadoc, they will ignore it and use it because it exists
(rgoulding,
16:56:42)
- it implies it is "approved" (rgoulding,
16:56:47)
- skitt in genius/netvirt you are firing off so
many small transactions that it is nearly hard to recover
anyway (rgoulding,
16:57:16)
- vorburger thinks that it will just make the
codelook ugly instead of actually influencing people to do it
correctly (rgoulding,
16:57:47)
- vorburger is there a middle ground by
dismissing the MoreFutures facade (which hides some of this)
(rgoulding,
16:58:48)
- skitt feels like that this is a waste
oftime (rgoulding,
16:59:09)
- vorburger the code to add a callback is like 5
lines instead of one utility line (rgoulding,
16:59:28)
Meeting ended at 17:02:26 UTC
(full logs).
Action items
- Sai and Luis to let Atul Gosain know the NETCONF password encryption patch is being reverted to unblock the Carbon-SR1 release
- to test the encryption mechanism in master
- tcere to +2 netconf encryption patch andemail releng team to merge the patch
- vorburger to open a bug against releng/builder for wiping sites
- ecelgp to push karaf.log or thread dump depending on whetherCPU spikes to 100%
Action items, by person
- vorburger
- vorburger to open a bug against releng/builder for wiping sites
- UNASSIGNED
- Sai and Luis to let Atul Gosain know the NETCONF password encryption patch is being reverted to unblock the Carbon-SR1 release
- to test the encryption mechanism in master
- tcere to +2 netconf encryption patch andemail releng team to merge the patch
- ecelgp to push karaf.log or thread dump depending on whetherCPU spikes to 100%
People present (lines said)
- rgoulding (91)
- odl_meetbot (4)
- zxiiro (3)
- vorburger (1)
Generated by MeetBot 0.1.4.