14:25:57 #startmeeting 14:25:57 Meeting started Mon Jun 20 14:25:57 2011 UTC. The chair is pmuir. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:25:57 Useful Commands: #action #agreed #help #info #idea #link #topic. 14:26:17 #topic trustin 14:26:29 back from one week pto 14:26:42 now catching up e-mail 14:27:05 I am going to keep working on EDG 14:27:13 .. this week. 14:27:22 That's all for now. 14:28:00 trustin: :-), ok i'll go next 14:28:12 unless anyone has any comments/questions? 14:28:37 pmuir: no questions :) 14:28:44 #topic pmuir 14:29:04 Last week I was working on the redo of classloading 14:29:42 I have done pretty much all of it now - so that everything uses a defined classloader rather than the TCCL 14:30:15 by default the TCCL is looked up when a GlobalConfiguration or Configuration is created 14:30:32 and that is the classloader used for all of the components that need classloader access 14:31:12 (all components are either CacheManager scoped or NamedCache scoped, that is what decides whether to use the ClassLoader from the GlobalConfiguration or Configuration) 14:32:16 if you load a configuration from disk either by default, or by passing in a file name, you don't get the option of specifying a classloader (obviously we have to use a classloader to load the config file) 14:32:52 otherwise, when you create the config (e.g. an empty ctor or specifying an InputStream or similar) then you also can specify a classloader 14:33:08 I think this is working ok, but will be interested in real feedback on it 14:33:41 the main thing I see being an issue is that I have changed the semantics by default from "always use the current TCCL" to "use the TCCL from whene the cache/cachemanager was set up" 14:33:44 which is a bit different 14:34:23 I also need to work the new classloading stuff into the marshaller framework - the API (cache.with(myClassLoader).get()/put() etc.) is in place, but it's not wired in yet 14:34:59 this is primarily due to (a) the DI in Infinispan not being designed to cope with anything finer grained than named caches and (b) the marshaller currently not being isolated to a named cache 14:35:15 pmuir trustin vblagoje: brb, got a surprise visit from some guys to estimate my flat renovation 14:35:21 I've also been writing up docs + blogs on the group api 14:35:37 dberindei: np, just say when you are back and we can catch up with you :-) 14:35:48 which are posted, and also on vnodes (not posted yet) 14:36:17 this week I plan to finish up the classloading stuff (add it to the marshaller) + add some tests + docs for it 14:36:52 i'm also going to take over migrating the docs to confluence from vblagoje (hopefully this week, need to chat to the .org team) which I think will be a big improvement for us 14:37:23 and start to take a look at some of the examples and tools which QE have been working on, see if we can't integrate those into the OSS project 14:37:25 pmuir, what about migration to confluence, can your docs wait and/or be written for confluence right away? 14:37:34 :- 14:38:14 you got this email from vlastimil? they are basically ready to start the migration 14:38:35 vblagoje: yeah, we can migrate when ready I think 14:38:50 thing to do is just to set a date after which people can no longer edit in sbs 14:39:00 ok pmuir 14:39:34 vblagoje: i suggest we do that this week, shall we say thursday? 14:39:54 dberindei: trustin: this will affect you too :-) 14:40:19 no problem :-) 14:40:37 vblagoje: btw I assume that manik told you that he asked me to take this off you as he guessed you were quite busy ;-) 14:40:53 we are moving only the user guide pages, aren't we? 14:41:14 yes, I'd appreciate it pmuir, I am working on JGroups features now, galder is on vacation and sanne as well I think. so noone is working on docs except you 14:41:29 which is good because this week is ideal for a move then 14:41:42 #action pmuir set up migration of user guide from sbs to confluence for thursday 23rd june with .org team 14:41:57 I don't think I will spend time of ispn documentation this week either. 14:42:23 s/of/on/ 14:42:48 pmuir: I'm back 14:42:59 #action pmuir coordinate and get sign off from team post migration to ensure all content migrated properly - aim to have this complete by wednesday 29th June 14:43:26 ok, and I will shut down edits to confluence to just core team until a full review is done to confirm content was migrated properly I think 14:43:46 and i think that will keep me busy this week :-) 14:43:58 yeah for sure 14:44:01 any other thoughts re doc migration or classloading? 14:44:32 if not i guess dberindei should go next whilst we have him here :-) 14:44:36 #topic dberindei 14:45:29 I was going to ask about classloading, can't we put the classloader in the InvocationContext? or is that what you're doing? 14:47:15 anyway, last week I worked on rehashing issues again 14:47:37 specifically ISPN-1012 14:47:38 jira [3ISPN-1012] With L1 disabled a new node join may result some entries are not invalidated for non-owner in DIST mode [10Resolved (Done) Bug,7 Critical,6 Dan Berindei] https://issues.jboss.org/browse/ISPN-1012 14:48:16 and after that I tried to ensure the state from other nodes is applied while holding the transaction lock 14:49:43 dberindei: for the classloading, you mean to pass the classloader to the marshaller? yes, that could work, I will look at it 14:50:00 pmuir: yes 14:50:13 dberindei: thanks for the hint :-D 14:50:14 I tried to move the applyState call to the rehasher thread, but that proved too complex because it's hard to know when we have received all the updates from the other threads 14:51:19 so what I did instead was to ignore the transaction lock while applying state, by making all commands with Flag.SKIP_LOCKING also ignore the transaction lock 14:52:35 I still wait in the rehasher thread until I have confirmation from the coordinator that all nodes have pushed state before I release the transaction lock 14:53:12 that seems to work really well in the unit tests, I hope it's not going to be a problem with very large clusters... 14:55:36 I'm now looking at the remaining errors in the test suite 14:57:34 that includes the failure that sannegrinovero reported on the mailing list, btw 14:57:54 pmuir vblagoje trustin: that's it from me 14:58:12 ok i can go next 14:58:23 #topic vblagoje 14:58:30 ok vblagoje :-) 14:58:37 i finished work on ISPN-1160 14:58:38 jira [3ISPN-1160] fetchInMemoryState doesn't work without FLUSH protocol for udp [10Resolved (Done) Bug,7 Major,6 Vladimir Blagojevic] https://issues.jboss.org/browse/ISPN-1160 14:59:15 which was a major pain because it made our udp stack practically unusable 14:59:30 and I was looking for a elegant solution 14:59:44 which also turned out to be a simple one as well 15:00:08 as far as I am concerned ISPN-83 is still just formally opened 15:00:09 jira [3ISPN-83] Remove dependency on JGroups FLUSH [10Reopened (Unresolved) Task,7 Critical,6 Vladimir Blagojevic] https://issues.jboss.org/browse/ISPN-83 15:00:15 and will close it this week 15:00:56 I did not work on other items as I have returned to complete some oustanding features on JGRoups 15:01:13 so thats it from me 15:03:26 vblagoje: ok thanks 15:03:37 any comments for vblagoje dberindei trustin? 15:03:50 none from me 15:04:27 none from me either 15:10:48 dberindei pmuir trustin vblagoje hi guys - I'm back. Did you guys capture meeting minutes on MeetBot? 15:11:04 yeah we were about to end the meeting 15:11:08 #endmeeting