On Mon, Sep 4, 2017 at 4:53 PM, Jozef Bacigál <jozef.baci...@pantheon.tech> wrote:
> Hi Michael, > > > > It seems to be the same issue over and over. What I mean it is always > issue with transaction chain. Do we need so many bugs opened for it? Let’s > keep it simple as possible because we are lost in this many bugs and it may > lead to confusion :) > FYI I've just attached a brand new trace:transactions output obtained via the new Bug 9060 tooling to https://bugs.opendaylight.org/show_bug.cgi?id=9096 which others may find interesting - and will shortly be opening even more bugs for what's in there, for other projects, to create even more general confusion! :) But yes please do feel absolutely free to just go ahead and close https://bugs.opendaylight.org/show_bug.cgi?id=9101 as a duplicate of https://bugs.opendaylight.org/show_bug.cgi?id=9038 in openflowplugin, if there is no point / no new additional information of value to you in it, and you are sure that it's about the same thing as 9038 and what you're doing there will fix this - not a problem at all. Likewise https://bugs.opendaylight.org/show_bug.cgi?id=9071 - if you have one fix for all of these 3, please just close and duplicate as you see fit, and have 1 (9038) instead of 3 (9038 + 9071 & 9101). But the https://bugs.opendaylight.org/show_bug.cgi?id=9034 is the one that just says "we have an OOM problem, due to TransactionChain(s!) not being closed... somewhere", and whether 9038 (=9071/9101) really is the fix for that, or indeed is one of more required fixes like that required in other projects, is something we IMHO should confirm (once we have your fix - thank you!), so I would like to keep that one open (in controller), even once you close 9038/9101 (in openflowplugin) ... like for example, based on what I've just attached go Bug 9096, I really have no way of knowing whether your openflowplugin bug 9038/9101/9071 or the ovsdb bug 9072 or 9073 are the real culprit - or am I missing some way by which we could tell? As for the other "so many bugs" I've opened today, and perhaps more to come, they are for other issues - for WriteOnlyTransaction or ReadOnlyTransaction but not TransactionChain, and in other projects. My understanding is that those also lead to memory leaks (but different from Bug 9034, understood). If that is wrong and a waste of time, then please someone shout STOP! ;-) > I already working on patch for all unclosed and/or illegalState txChains > in all OFP. I will let you know when it is ready to test it. > Wonderful - thank you so much! Thanks > > > > Jozef > > > > *From:* Michael Vorburger [mailto:vorbur...@redhat.com] > *Sent:* Monday, September 4, 2017 4:44 PM > *To:* Robert Varga <n...@hq.sk>; Tomáš Slušný <tomas.slu...@pantheon.tech>; > Jozef Bacigál <jozef.baci...@pantheon.tech> > *Cc:* Abhijit Kumbhare <abhijitk...@gmail.com>; openflowplugin-dev@lists. > opendaylight.org; controller-dev@lists.opendaylight.org; > mdsal-...@lists.opendaylight.org > *Subject:* Re: [mdsal-dev] [openflowplugin-dev] Fwd: Bug 9038 - > IllegalStateException: Attempted to close chain with outstanding > transaction PingPongTransaction at org.opendaylight. > openflowplugin.impl.device.TransactionChainManager.createTxChain > > > > On Thu, Aug 24, 2017 at 1:10 PM, Robert Varga <n...@hq.sk> wrote: > > On 24/08/17 09:39, Tomáš Slušný wrote: > > > > Hello Michael, > > > > so, according to stack trace, it looks like OpenFlowPlugin transaction > > chain manager got notification that transaction chain failed, and what > > we are doing then are that we create new transaction chain and close the > > failed transaction chain. But it looks like in time we are closing the > > failed transaction chain, there is still ping pong transaction open > > (what, according to logs, is probably the one who triggered the failure > > of chain and also the recreation of the chain), so I am not sure if this > > is really bug in OpenFlowPlugin. Adding controller and mdsal to cc. > > OFP has a currently-open transaction when from the failed transaction > chain (i.e. between allocate and submit/cancel) when it is calling > close(). This points towards missing synchronization between the > callback code and the code that is using the chain. > > https://bugs.opendaylight.org/show_bug.cgi?id=9101 may interest you as > well in this context? > > Bye, > Robert > > > _______________________________________________ > mdsal-dev mailing list > mdsal-...@lists.opendaylight.org > https://lists.opendaylight.org/mailman/listinfo/mdsal-dev > > >
_______________________________________________ controller-dev mailing list controller-dev@lists.opendaylight.org https://lists.opendaylight.org/mailman/listinfo/controller-dev
