14:03:44 #startmeeting 14:03:44 Meeting started Mon Jun 13 14:03:44 2011 UTC. The chair is manik1. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:03:44 Useful Commands: #action #agreed #help #info #idea #link #topic. 14:03:58 Greetings all 14:04:07 weekly catchup time 14:04:19 mmarkus: want to go first since you're here? 14:04:36 dberindei: feel free to reopen and assign it to me 14:04:47 galderz1: ok 14:04:57 manik1: sure 14:05:13 hi all 14:05:28 last week I've mainly worked on improving RG to support transactions benchmarking 14:05:46 then run some benchmarks with different transaction configurations 14:06:18 I'm writing a blog about that right now, will send it shortly for review 14:06:32 galderz1: done 14:06:49 also started working on my jazoon preso, which I need to finish by Wen 14:07:15 mmarkus: when are you speaking at Jazoon? 14:07:34 next Wen 14:07:51 #info mmarkus at Jazoon on Weds, 22nd June 14:08:02 mmarkus: ok 14:08:24 mmarkus: What else for this week? Any other 5.0 issues that need attention? 14:09:27 I still have https://issues.jboss.org/browse/ISPN-842 14:09:29 jira [3ISPN-842] FirstAvailable load balance policy for Hot Rod clients and extend reach of load balance policies [10Open (Unresolved) Feature Request,7 Major,6 Mircea Markus] https://issues.jboss.org/browse/ISPN-842 14:10:03 That's just adding another LB policy to the HR client? 14:10:58 yes 14:11:04 there was something else, one sec 14:13:11 https://issues.jboss.org/browse/ISPN-1175 14:13:12 jira [3ISPN-1175] XA transaction rollback failure throws the wrong exception and fails to cleanup the transaction table [10Open (Unresolved) Bug,7 Major,6 Mircea Markus] https://issues.jboss.org/browse/ISPN-1175 14:14:42 also another one I'm about to raise around transactions and node failures, which Dan and I spotted last week during a code walk through 14:15:04 mmarkus: ok. Targetting this week for closing these? 14:15:50 manik1: yes, for sure 14:17:09 just to let you know, I'll be off to Zurich Wen-Thu next week. I'll also send an email around 14:17:09 mmarkus: great. 14:17:19 mmarkus: ok 14:17:24 Next? 14:17:28 vblagoje? 14:17:34 ah, and thanks everybody for the feedback on transactions improvements! 14:17:46 mmarkus: np. :-) 14:18:24 vblagoje must be getting coffee. :-) dberindei: want to go next? 14:19:18 manik1: ok, I spent most time last week working on ISPN-1012 14:19:19 jira [3ISPN-1012] With L1 disabled a new node join may result some entries are not invalidated for non-owner in DIST mode [10Reopened (Unresolved) Bug,7 Critical,6 Dan Berindei] https://issues.jboss.org/browse/ISPN-1012 14:19:23 #topic dberindei 14:19:40 ok, how's it looking? 14:20:13 i am on the phone,, give me 5 min 14:20:31 it's working now, in the end I had changed RebalanceTask to invoke the InvalidateL1Command through the interceptor chain 14:20:58 and I also had to change DistLockingInterceptor to allow removals regardless of locality 14:21:27 dberindei: ok. What else? Any other issues regarding rehashing? 14:21:35 Stability of rehash-related tests? 14:21:58 I still see some failures, that's why I haven't created the pull request yet 14:22:54 I have also fixed an embarrasing problem in my ISPN-1092 commit, I had changed the algo by mistake and I didn't run the hotrod client suite 14:22:55 jira [3ISPN-1092] TopologyAwareConsistentHash with virtual nodes enabled can return two owners for a key when there is only one node in the cluster [10Resolved (Done) Bug,7 Minor,6 Dan Berindei] https://issues.jboss.org/browse/ISPN-1092 14:23:19 luckily TopologyAwareStateTransferTest was failing because of it and pmuir noticed it :) 14:23:30 dberindei: ok! :-) 14:23:50 dberindei: I presume you're still pressing on with this this week? 14:24:09 I'm creating a pull request right now for ISPN-1171 and then I'm back to the rehash tests 14:24:10 jira [3ISPN-1171] State transfer should not write anything to the shared cache store or clear any cache store [10Open (Unresolved) Bug,7 Major,6 Dan Berindei] https://issues.jboss.org/browse/ISPN-1171 14:25:05 so yeah, I'll probably be busy the rest of the week :) 14:25:37 dberindei: ok. How about overall testsuite stability? 14:25:47 Have you been able to re-enable tests on ISPN-1123? 14:25:48 jira [3ISPN-1123] Stabilise test suite [10Open (Unresolved) Task,7 Blocker,6 Dan Berindei] https://issues.jboss.org/browse/ISPN-1123 14:27:12 who's up next, btw? sannegrinovero, vblagoje, pmuir, galderz1 ? 14:27:18 i can go 14:27:21 ok 14:27:24 not yet, but I did find a problem with KeyAffinityServiceImpl not shutting down properly all the time 14:27:26 #topic galderz1 14:27:54 last week I was integrating 2LC with transaction synchonizations - it all looked good but seems like the transaction manager differences have an impact 14:27:56 dberindei: ok - feel free to ping others for help there, since some of the tests pertain to others here 14:27:57 as scott discovered 14:28:14 sure manik1 14:28:18 galderz1: you mean the AS7 integration? 14:28:23 yeah 14:28:50 that should be under control now with scott's fixes, so nothing urgent to do in 2LC 14:28:58 i also had some git nightmares, prob something to do with github, but drove me mad 14:29:09 apart from that, several CR5 jiras 14:30:16 btw, i'm off starting this Friday until the 27th inclusive 14:30:24 galderz1 anything I can help with re:Synchronization ? 14:30:46 mmarkus: no, it's fine - the 2LC testsuite needs migrating to JBoss TS though 14:31:05 galderz1: ok, cool. In the meanwhile if you have some time this week could you help dberindei with ISPN-1123? As well as the issue with the war deployment in AS 7 (assigned to trustin) ? 14:31:06 jira [3ISPN-1123] Stabilise test suite [10Open (Unresolved) Task,7 Blocker,6 Dan Berindei] https://issues.jboss.org/browse/ISPN-1123 14:31:17 Hibernate 4 has already done so - so i need to see what these guys have done to make sure the database drive, h2, hooks up nicely with the tx manager 14:31:31 sure manik1 14:32:18 galderz1: thanks 14:32:27 who's next? pmuir? sannegrinovero? vblagoje? 14:32:35 yeah, i can go 14:32:39 #topic pmuir 14:32:49 ok 14:33:11 been working on making the classloader used to load classes and resources explicitly configurable 14:33:20 we now have three classloading "scopes" 14:33:39 1) CacheManager, and you can configure the classloader in use using GlobalConfiguration 14:34:01 2) Cache, and you can configure the classloader in Configuration 14:34:12 it will use the one from CacheManager by default 14:35:07 3) and then per cache "use" (which is the much discussed Cache.with(ClassLoader) option) 14:35:35 I think most of the rewire is done, just finishing that up now 14:36:27 (the TCCL is still used as the default CL to use, it's just configurable) 14:36:32 pmuir: can you configure a classloader in the xml? 14:36:37 dberindei: no 14:36:45 dberindei: idk how you would do that? 14:37:07 i mean, i thought about this but couldn't see a sane way to allow it 14:37:36 pmuir: agree, I just saw "configurable" and I immediately thought about the xml config :) 14:37:43 as there is no metamodel for classloaders to specify them declaratively 14:37:51 dberindei: basically you do 14:37:58 new GlobalConfiguration(myCl); 14:39:00 I think this will work fine for the usecases where it is needed 14:39:28 pmuir: good stuff. 14:39:50 I plan to finish this up this afternoon 14:39:55 pmuir: yeah, I can only see it being useful to folks who cfg stuff programmatically anyway 14:40:01 tomorrow I am in Paris for the JUG 14:40:07 pmuir: great, ping me when you're done and I'll cut another CR 14:40:14 so the next 24h will be prep for that 14:40:30 and then I want to spend some time documenting the group api and adding the xml config for it 14:40:41 and I have a few bugs to fix which will take me to friday I think 14:40:43 pmuir: any more feedback/info/benchmarks on vnodes? 14:41:04 i think with the change that dberindei just did and i merged this am 14:41:16 we should actually see something in the benchmark space that makes sense 14:41:26 and am assuming that dberindei has this on his todo? 14:41:55 and I will spend some more time documenting/blogging on them as well 14:42:09 manik1: no benchmark from me yet, but I see mmarkus got radargun working on the test cluster so I'll get his help and run the vnodes benchmark there too 14:42:19 dberindei: ok, cool. 14:42:47 sanne / vblagoje: up next? 14:42:54 manik1, sure 14:42:55 ok I'll go 14:43:02 vblagoje, goes first :) 14:43:04 #topic vblagoje 14:43:10 ok 14:43:39 i was pulling my hair for a whole week around ISPN-1160 14:43:40 jira [3ISPN-1160] fetchInMemoryState doesn't work without FLUSH protocol for udp [10Open (Unresolved) Bug,7 Major,6 Vladimir Blagojevic] https://issues.jboss.org/browse/ISPN-1160 14:44:16 as manik suggested I tried to make it work with invokeInFuture but no success 14:44:37 there is a solution with new Thread which is not perfect and it works 14:45:01 but I would still like to find a more elegant one 14:45:42 I am not sure what to do. should we put in one that works and revisit? i could use some help and review there 14:46:21 so if anyone is volunteering to help i can direct you to branches on my fork 14:46:28 vblagoje: I thought the STREAMING_STATE_TFR workaround did the trick? 14:47:08 it did, but many people have udp stack without that switch 14:47:21 and they will get stuck and will have a bad user experience 14:47:36 vblagoje: So the better fix then, IMO, is to change the default cfgs we have to set use_default_transport="true" 14:47:54 true, i thought about that, but what about existing configs out there? 14:48:00 or even to do it dynamically in the JGroupsTransport before starting up (unless this is disabled by a user flag) 14:48:23 vblagoje: I think that is far cleaner than a separate thread/executor. 14:48:31 Lets chat about this offline… 14:48:38 ok manik1 lets chat 14:48:46 vblagoje: Anything else? What else have you got on your plate? 14:49:01 Wasnt there something about some tests not being run in the testsuite? 14:49:14 i have ISPN-83 which I think is non-existent any more, it is just that i did not close it yet 14:49:15 jira [3ISPN-83] Remove dependency on JGroups FLUSH [10Reopened (Unresolved) Task,7 Critical,6 Vladimir Blagojevic] https://issues.jboss.org/browse/ISPN-83 14:49:28 i have no fix for it and no pull 14:49:43 but it seems to be gone 14:50:41 the original test michal used to raise, passes now and I have not seen it raised 14:51:07 other than that everything else is not a big priority 14:51:12 so much from me 14:52:20 vblagoje: ok, so have you gone ahead and removed FLUSH from the configs we ship? 14:52:41 yeah, they have been removed long time ago 14:53:04 we removed it in january 14:53:30 vblagoje: just double-check that they weren't re-added, to be removed only after ISPN-83 was complete 14:53:31 jira [3ISPN-83] Remove dependency on JGroups FLUSH [10Reopened (Unresolved) Task,7 Critical,6 Vladimir Blagojevic] https://issues.jboss.org/browse/ISPN-83 14:54:22 no, no we put them back in on 4.2 branch and we kept 5.0 clean from FLUSH since January 14:54:46 vblagoje: are you going to then look at ISPN-1108 this week? 14:54:47 jira [3ISPN-1108] Test(s) ignored by maven [10Open (Unresolved) Bug,7 Blocker,6 Vladimir Blagojevic] https://issues.jboss.org/browse/ISPN-1108 14:54:51 vblagoje: ok 14:56:05 yes manik as soon as this current "nightmare" is behind me 14:56:10 vblagoje: :) ok 14:56:19 #topic sanne 14:56:23 you next, sanne ? 14:56:31 manik1, yes 14:56:46 So, I've been mostly reviewing code, both pull requests from all of you and several contributions 14:56:58 my main contribution to Infinispan this week was around the issues with passivation reported via a Hibernate Search users. 14:57:34 Did some small patches around ISPN-1123, as I'm getting close to be able to test infinispan/core. 14:57:35 jira [3ISPN-1123] Stabilise test suite [10Open (Unresolved) Task,7 Blocker,6 Dan Berindei] https://issues.jboss.org/browse/ISPN-1123 14:57:41 Sorry, I could not find time to work on big tasks, nor review the C code; 14:57:53 sanne: np. Are these on your list for this week? 14:58:18 or are you busy on OGM/query parsing this week? 14:58:20 manik1, no, I'm overbooked, unless you need that ? 14:58:30 also I should recover the locking discussions (sorry Mircea!), but I got too much out of context from that currently. 14:59:08 I couldn't verify if indeed my point 5) which I had added to the doc is redundant (we explain in quite different ways); if you think it gets people confused, mmarkus feel free to delete it for now, I recover it later to think about that. 14:59:31 Also I've been working on my new secret project, hope we can reveal it soon: it's called Infinispan 5.1 14:59:48 I'm basically withholding experimental patches for 5.1; they're too experimental yet anyway to be merged anywhere. 15:00:05 especially ISPN-200 is making some progress, and I'm going to update the Lucene Directory basic structure. 15:00:07 jira [3ISPN-200] Distributed queries [10Open (Unresolved) Feature Request,7 Major,6 Sanne Grinovero] https://issues.jboss.org/browse/ISPN-200 15:00:34 sanne: +1. People tend to comment around it which is a bit sitracting. I'll remove it and then we can have a chat whenever you feel like 15:00:52 sanne: not super-urgent, although I would like to get to it reasonably soon. 15:00:56 In both cases I'm having trouble as Lucene is making progress very fast, and I'm not able to follow all commits every day; both these tasks are affected by the new API changes and I should take some time to provide feedback to them as well, before we get stuck with an incompatible API 15:01:41 I'm going to need several days off next week (still have to ask for them), and focusing on OGM this week, so I won't be helpful on Infinispan in the next 2w. Also because I'm leaving, I need to provide parser updates to emmanuel or he and cloud-tm will be blocked. 15:01:42 are you using distributed execution for ISPN-200 sanne? 15:01:43 jira [3ISPN-200] Distributed queries [10Open (Unresolved) Feature Request,7 Major,6 Sanne Grinovero] https://issues.jboss.org/browse/ISPN-200 15:01:47 sanne: is your new "secret project" related to the locking changes? :) 15:02:59 no manik1, not had time to really experiment with it. but at some point I'll have to start the new design of the lucene index, using the new codecs API coming in Lucene 4, and yes I'm trying to understand more on locking because of that. 15:03:17 sanne: ok, cool. 15:03:55 Ok, anyway we're out of time, thanks for the updates guys. Happy hacking. 15:04:04 also, last week helped the cloud-tm partners to have a Ruby application using Torquebox using HibernateOGM / Infinispan as storage engine 15:04:17 that's all manik1 15:04:23 Good stuff, sanne 15:04:33 #endmeeting