13:02:50 <gsmet> #startmeeting
13:02:50 <jbott> Meeting started Tue May  8 13:02:50 2018 UTC.  The chair is gsmet. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:02:50 <jbott> Useful Commands: #action #agreed #help #info #idea #link #topic.
13:02:51 <jbossbot> Title: MeetBot - Debian Wiki
13:02:56 <gsmet> yeah, it works better :)
13:03:18 <DavideD> So, the main work I've done was related to PR review
13:03:22 <gsmet> fax4ever: thanks for warning me
13:03:26 <gsmet> #topic Progress Davide
13:03:31 <fax4ever> you're welcome
13:03:47 <DavideD> Mainly worked on PR review
13:04:14 <DavideD> They are getting better ad better but at the same time they are a bit fragile
13:04:50 <DavideD> We had several improvements around Neo4j native queries with parameters and yesterday we received a PR for the creation of the indexes
13:05:46 <DavideD> Sergey sent the pull request about the pool for Neo4j embedded that should improve performance
13:06:25 <DavideD> It seems to cut down the time of the build by not restarting Neo4j every time but it highlights some problems in our test suite
13:06:36 <DavideD> so I'm not sure if I'm going to include it
13:07:09 <DavideD> fax4ever, seems able to replicate the problem consistently, maybe when you finished the other issue you can have a look at it?
13:07:28 <fax4ever> sure!
13:07:30 <gsmet> DavideD: the issue looks a lot like the Travis issue we had (or still have?)
13:07:40 <DavideD> Yeah, I think it's exactly the same
13:07:41 <gsmet> it would be nice if we could fix it once and for all if we can reproduce it
13:08:00 <DavideD> Infact in Travis it happens when we use a remote server and therefore we don't recreate the db every time
13:08:05 <gsmet> I'm +1 to include the PR if it allows to fix test issues and be more safe on the test side
13:08:13 <DavideD> Similar to what's happening now
13:08:36 <gsmet> (not saying it's going to be easy to spot the culprit...)
13:08:38 <DavideD> Also the pr sergey sent cut almost a minute from the build time
13:09:04 <DavideD> To be honest I don't think we will figure it out... good luck fabio :)
13:09:11 <fax4ever> maybe we can improve the log of error diplaying also the content of the cache
13:09:29 <DavideD> Feel free to experiment with it
13:09:31 <fax4ever> this could help to reach the test that does not clean
13:09:34 <gsmet> yes it's doable, I did that a long time ago when I was looking for something
13:09:42 <gsmet> I say it's all related to Luke and Leia
13:09:51 <gsmet> all our test issues are always about those two
13:10:05 <DavideD> Sure :) let's update the reference to more modern times
13:10:13 <DavideD> Next test I want to see the Avengers
13:10:21 <fax4ever> ah ok! ideally each test should clean up itself. right?
13:10:34 <DavideD> We can talk about it later Fabio
13:10:38 <fax4ever> +1
13:10:39 <gsmet> fax4ever: I think what you're suggesting is worth it, it might help to narrow the issue down
13:10:57 <DavideD> But yeah, Im sure the tests clean themselves because they work when you run them separately
13:11:03 <sannegrinovero> no spoilers in the tests please ;)
13:11:11 <DavideD> ann most of the tests are run on all the dialects  ...
13:11:14 <gsmet> fax4ever: then you could find the right one by changing a bit the tests using these entities (e.g. having Leia1 and Luke 23)
13:11:37 <fax4ever> ok
13:11:49 <DavideD> In theory yes, but I suspect the test fails because some other test before hasn't delete something
13:12:05 <DavideD> You will need to tweak the code a bit for sure
13:12:07 <gsmet> yes or there's some hidden failures somewhere
13:12:19 <DavideD> yes, that could be as well
13:12:40 <gsmet> anyway it would be nice to have Travis running again so if we can find the culprit, that would be nice
13:12:50 <sannegrinovero> DavideD, let's move on from details, this was discussed on gh already?
13:13:20 <DavideD> Yeah, but apparently it attracts the curiosity of the people present :)
13:13:43 <DavideD> Anyway,  also started the post about feature packs but didn't pregress much
13:14:12 <DavideD> I should be able to complete it and show you guys something  this week this week
13:15:00 <DavideD> By the way
13:15:26 <DavideD> I'm kind of happy about the user questions on SO, it seems that most of them were able to solve the issues they had
13:15:34 <DavideD> thanks to our suggestions
13:16:05 <DavideD> I guess, next tow weeks?
13:16:15 <gsmet> #topic Next 2 weeks Davide
13:16:45 <DavideD> Mainly work on the article and finally tackle ORM 6
13:17:03 <gsmet> \o/
13:17:09 <DavideD> I should be able to work on them in this iteration
13:17:43 <DavideD> Also need ot update NEo4j on CI and update one of the db ORM uses for test
13:18:15 <DavideD> Chagne the type of the machine because Amazon isn't going to support the current type anymore
13:18:32 <DavideD> I think that's all
13:18:43 <gsmet> OK, cool thanks
13:18:51 <gsmet> #topic Progress Fabio
13:18:56 <fax4ever> Thank you
13:19:07 <fax4ever> I started to working on upgrade to Hibernate ORM 5.3.0.CR2
13:19:25 <fax4ever> From the 5.3.0.CR1 to 5.3.0.CR2 it has been changed the way Hibernate ORM handles the reuse of transaction instances,
13:19:34 <fax4ever> over the same session.
13:19:43 <fax4ever> Andrea B. helped me to understand the inner workings.
13:19:54 <fax4ever> In CR1, if the instance was invalid after the first use, Hibernate ORM will detect them and recreate it,
13:20:03 <fax4ever> in CR2 it is no longer true.
13:20:14 <fax4ever> Invalidation of transaction after commit seems an old behavior,
13:20:23 <fax4ever> Andrea B. is currently working to remove it from Hibernate ORM.
13:20:33 <fax4ever> As you can see, fix has expected for 5.3.0.Final.
13:20:45 <fax4ever> Furthermore, in the last two weeks, I applied some changes on Infinispan Remote query support issue.
13:20:54 <fax4ever> Suggested by Davide
13:21:19 <fax4ever> I worked in order to remove some limitation, for instance now we support Boolean or Char as query parameters.
13:21:30 <fax4ever> I included in the PR some refactorings, to avoid code duplication and improve reuse,
13:21:40 <fax4ever> that has affected Hibernate OGM core, Neo4j and MongoDB modules too.
13:21:52 <fax4ever> Also I enabled support for native query for Infinispan remote.
13:21:53 <DavideD> fax4ever, remember to keep in touch with Andrea and try to test the fix before they release it
13:22:02 <fax4ever> +1
13:22:25 <fax4ever> In the end I had 2 days of National holiday.
13:22:34 <fax4ever> This is all about my last two weeks. Thank you.
13:22:39 <gsmet> thanks
13:22:46 <gsmet> #topic Next 2 weeks Fabio
13:22:57 <fax4ever> For the next two weeks...
13:23:18 <fax4ever> I'm going to apply the last updates suggested by Davide and Guillaume to Infinispan remote query issue.
13:23:27 <fax4ever> I believe we could finally resolve it in few days.
13:23:40 <fax4ever> Then I will review the new EPAM pull request, that includes support for Infinispan remote task execution.
13:24:00 <fax4ever> When both Query and Remote Task issues will be merged on the master of Hibernate OGM, maybe we should release
13:24:01 <fax4ever> the Alpha2. WDYT?
13:24:22 <fax4ever> When the Alpha2 will be released, I would like to update the OpenShift Hibernate demo,
13:24:33 <fax4ever> adding to it an example of query and an example of stored procedure for the OGM / Infinispan part.
13:24:48 <fax4ever> Besides I would like to working more on MongoDB and Neo4j dialects, solving more issues and reviewing many PRs on them.
13:25:03 <fax4ever> In general I would like to do whatever could help our projects and the team.
13:25:17 <fax4ever> It's all about my next two weeks. Thank you!
13:25:24 <gsmet> thanks Fabio
13:25:30 <sannegrinovero> yes we'll need to evolve the demo further
13:25:41 <gsmet> #topic Progress Guillaume
13:25:42 <sannegrinovero> I need Search and ES to be part of it :)
13:25:54 <fax4ever> ;)
13:26:06 <gsmet> I committed a few things to Search 6, mostly new tests (and related fixes)
13:26:19 <gsmet> reviewed a couple of PRs on 6 too
13:26:41 <gsmet> I did my fair share of reviews/design discussions on HV
13:27:07 <gsmet> Marko is working on the JSON support so I spent some time trying to build a general design for the feature
13:27:29 <gsmet> I also prepared the 6.0.10 release, we will release soon
13:27:50 <gsmet> it fixes an annoying bug with CDI that could affect WF/EAP users so I would like to upgrade WF with it
13:28:08 <gsmet> I released the BV TCK 2.0.3 version
13:28:31 <gsmet> the difficult thing is that we will need the TCK tests and the HV version together at the same time for WF
13:28:41 <gsmet> as we had to change some TCK tests and the implementation
13:29:01 <gsmet> I also worked on this branch filtering thing for Jenkins
13:29:06 <gsmet> apparently, it's working well
13:29:24 <gsmet> I plan to submit a PR soon, not sure it will get merged but I'll try
13:29:55 <gsmet> and finally, I spend some time analyzing this memory dump for ORM
13:30:08 <gsmet> that's pretty much it for my last 2 weeks
13:30:17 <gsmet> #topic Next 2 weeks Guillaume
13:30:44 <gsmet> tomorrow and Friday (Thursday is a public holiday again...), I'll work on the ORM loader issue
13:30:57 <gsmet> I hope I'll get a prototype by then (and JIRA issues ;))
13:31:37 <sannegrinovero> we all love more JIRA issues ;)
13:31:48 <gsmet> I'll probably have some discussion with Marko about the JSON feature but he will be travelling for 2 weeks at the end of the week so it should be quiet on the HV front
13:32:11 <gsmet> I'll release HV 6.0.10, I need to get at least one more fix in
13:32:19 <gsmet> and the rest of the time, I'll work on Search 6
13:32:25 <sannegrinovero> how if HV doing in terms of JDK11 ?
13:32:33 <gsmet> didn't test it
13:32:42 <gsmet> JDK 10 is OK, apart from OSGi
13:32:56 <gsmet> 6.1 should be OK with JDK 11 as we removed JAXB
13:33:16 <gsmet> for 6.0, we need to check and write an article about how to run HV 6.0 on JDK 11
13:33:23 <sannegrinovero> right that's why I ask. HV is probably the project which is "closest" to have it done.
13:33:26 <gsmet> (not sure we will have 6.1 released on time)
13:33:45 <gsmet> yeah the issue is with all the tooling broken
13:33:53 <gsmet> it's not really easy to test JDK 11 right now
13:34:35 <gsmet> anyway, that's it for me
13:34:54 <gsmet> I'll plan some work on JDK 11 later this month
13:35:02 <gsmet> #topic Progress Sanne
13:35:34 <sannegrinovero> ok so I've tried to check various projects in terms of JDK10 and JDK11 compatibility
13:35:46 <sannegrinovero> especially as DavideD had installed them recently on ci..
13:36:01 <sannegrinovero> turns out much tooling is still not ready
13:36:13 <sannegrinovero> Maven seems to work now, provided you update some plugins
13:36:23 <gsmet> for JDK 11 too?
13:36:26 <sannegrinovero> Gradle "kinda works" if you use a nightly build
13:36:29 <sannegrinovero> yes JDK11 too
13:36:48 <gsmet> I was afraid about the -ea suffix as the patches I saw for 10 were expecting a dot
13:36:48 <sannegrinovero> however Karaf doesn't boot
13:37:10 <sannegrinovero> yes that's what breaks Karaf and Gradle 4.7
13:37:33 <sannegrinovero> but surefire was fixed (was it surefire? don't remember)
13:37:36 <gsmet> did you ping the pax-exam guys about it?
13:37:47 <gsmet> because I think their patch for 10 won't work for 11
13:37:52 <sannegrinovero> and we don't have a pax-exam working with karaf 4.2 either
13:38:01 <sannegrinovero> which means we can't run our OSGi tests on JDK10 either
13:38:18 <sannegrinovero> finally Byte Buddy needs an update
13:38:20 <gsmet> yeah they have a if (version.indexOf(".") > 0) {
13:38:32 <sannegrinovero> however if you update BB, the ORM enhancer is broken :)
13:38:49 <fax4ever> :P
13:38:58 <sannegrinovero> this is unexpected as I only did a micro update of BB, so I'll need to check how/why ORM breaks when updating it.
13:39:15 <sannegrinovero> but since Gradle doesn't work either it's a bit hard to do.
13:39:39 <sannegrinovero> so, in parallel with steve we where hardening our use of ForbiddenAPIs
13:39:56 <sannegrinovero> not least to avoid JDK methods which are being deprecated, so to make these updates smoother.
13:40:06 <sannegrinovero> this lead to some more polishing in both ORM and Search
13:40:24 <sannegrinovero> and Search finally got module names ! although just automatic-module-names
13:40:39 <sannegrinovero> we released an ORM CR and a Search CR
13:40:48 <sannegrinovero> (the search CR is to be announced in ~30 minutes)
13:41:00 <sannegrinovero> so we're experimenting now with WildFly 13 to update them
13:41:01 <gsmet> yup, I have a reminder for that :)
13:41:22 <sannegrinovero> now it turns out I might have some bad news about WildFly updates:
13:41:40 <sannegrinovero> we can't have two ORM versions in there, as we can have only one JPA API version
13:42:00 <sannegrinovero> so that means for WildFly to be certified it would need to be EE8 certified if we upgrade JPA22
13:42:15 <sannegrinovero> I'm still trying to understand - as I thought this was the plan all along
13:42:21 <gsmet> sannegrinovero: what's not ready for EE8?
13:42:29 <sannegrinovero> but it might be that it's too late for that, as EE8 cert is much more work
13:42:55 <sannegrinovero> gsmet, besides us still having some blockers open ? :P I don't know, I asked Jason if it's all in our hands or if there are other parties involved.
13:43:12 <gsmet> OK
13:43:21 <sannegrinovero> so for now.. let's keep working on this integration, it's not a wasted effort anyway
13:43:31 <sannegrinovero> worse case it goes in WF14 which is just 3 more months
13:43:57 <sannegrinovero> so that brings me to next 2 weeks smoothly:
13:44:08 <gsmet> #topic Next 2 weeks Sanne
13:44:28 <sannegrinovero> I'm in touch with both the perf team and the 2lc/clustering team
13:44:44 <sannegrinovero> both of them found some issues integrating latest ORM so now they have to be fixed quickly
13:44:59 <sannegrinovero> and we try to finish this integration, or decide it's too rushed.
13:45:18 <sannegrinovero> of course Search/OGM/Validator integration needs to be in the picture too
13:45:41 <sannegrinovero> so if there's feedback or needs for changes in ORM's SPI let's please figure it out ASAP
13:46:07 <sannegrinovero> also.. of course this implies the performance team might find more blockers
13:46:21 <sannegrinovero> that's just how it works, so no problem if we end up delaying
13:46:30 <sannegrinovero> but let's make sure we catch the train if it's possible
13:46:52 <sannegrinovero> (there's loads of community people commenting on WildFly forums/ jiras asking for the Hibernate stack to be updated)
13:47:29 <gsmet> yeah it's a bit unfortunate our competitors are using newer versions
13:47:29 <sannegrinovero> In parallel, we're discussing with some other teams how we can improve clustering of Search
13:47:36 <sannegrinovero> gsmet, +100
13:48:04 <sannegrinovero> especially we don't like our solutions based on JMS/JGroups, so for openshift would be nice to have better alternatives - beyond Elasticsearch of course.
13:48:21 <gsmet> ah too bad, I was going to say Elasticsearch :)
13:48:21 <sannegrinovero> as ES is nice but we don't want to put Openshift users in the lock-in position
13:48:31 <sannegrinovero> yea :)
13:48:40 <gsmet> do they have proposal?
13:48:55 <sannegrinovero> but also some people just want a little bit of trivial search, so not always ideal to start an ES cluster for that.
13:48:59 <gsmet> it might be good to know as maybe we shouldn't integrate the current clustering with Search 6
13:49:34 <gsmet> well, it's easier to set up an Es cluster than setup a JMS/JGroups cluster
13:49:44 <sannegrinovero> gsmet, I got no concrete proposal, but Emmanuel suggested that openshift might have an HA queue service which might be a nice replacement for JMS.
13:50:07 <fax4ever> nice
13:50:15 <gsmet> OK, so we would support clustering on OS + Elasticsearch for community?
13:50:29 <sannegrinovero> speaking of Emmanuel, is at Summit this week so I hope he can meet some of those guys in person and get some more concrete whiteboard going.
13:50:56 <sannegrinovero> gsmet, yes I think we'll always need both, and both solutions should be able to scale "OOB"
13:51:26 <sannegrinovero> I think that's all for me.
13:51:42 <gsmet> ok, thanks for sharing!
13:51:47 <sannegrinovero> if there are no questions I'l get back at bytebuddy debugging :/
13:51:51 <gsmet> let's close this meeting then
13:52:01 <DavideD> Thanks everybody
13:52:05 <sannegrinovero> thanks all!
13:52:06 <gsmet> and I'll push the announcement for 5.10.0.CR1
13:52:15 <gsmet> thanks everyone
13:52:19 <gsmet> #endmeeting