I don't believe 1159 is a bug if we assume the corresponding patch is supposed 
to fix something. It has no effect.

The filer may be expecting an expiration notification when the send thread 
terminates for reasons other than a SessionExpiredException having been thrown. 
One could distinguish between 'disconnected', 'expired' and 'terminated' -- 
though from my reading of the code, the latter should only occur if the user 
explicitly calls close() when the client is disconnected.

I do think we can tighten up state management in the client along the lines 
described in 1202. But this has more to do with close() and can be deferred to 
3.5.

Matthias


-----Original Message-----
From: Fournier, Camille F. [mailto:[email protected]] 
Sent: Thursday, September 29, 2011 8:34 AM
To: 'Mahadev Konar'; '[email protected]'
Cc: '[email protected]'; Spycher, Matthias; 'Benjamin Reed'
Subject: RE: Update on 3.4 rc.

I could really use some help on 1197 actually, if anyone is willing to help on 
the jira or ping me directly on this. Basically, that patch helps things quite 
a bit, but still doesn't fix the problem completely and I'm not sure where to 
go from here, if it is even possible to completely fix this issue. The 
alternative to the complexity in that patch is to just put a thread.sleep at 
the end of the command thread, before it closes the socket, to try and get data 
flushed out. Much simpler, still not perfect, but of course has other potential 
drawbacks. This is a major problem for anyone running servers in an WAN 
environment that needs to do monitoring, so it really should go into this 
release if at all possible.

If Mathias has some ideas on 1159 I'm all ears. I really don't feel comfortable 
throwing code in to "Fix" a concurrency bug we don't understand, and we don't 
have any explanation of the issue that makes sense yet. 


C

-----Original Message-----
From: Mahadev Konar [mailto:[email protected]]
Sent: Thursday, September 29, 2011 4:10 AM
To: [email protected]
Cc: [email protected]; [email protected]; Fournier, Camille F. [Tech]; Benjamin 
Reed
Subject: Update on 3.4 rc.

Hi all,
 We are very very close to getting a release candidate out. Here is an update 
on some work we need to do to finish off:

Patch Availables:

ZOOKEEPER-1197: Needs a reviewer. Pat/Ben would you be able to review?
ZOOKEEPER-1174: Already committed to 3.4. Is pending commit to 3.3 branch.
ZOOKEEPER-1181: I am reviewing this. Should be done by tommorrow.
ZOOKEEPER-1203: Looks like just needs to be marked fixed.
ZOOKEEPER-1206: Will check it in tommorrow morning as soon as hudson +1's.


Open Jiras:
ZOOKEEPER-1149: Pat again you are on the hook for this :).
ZOOKEEPER-1159: Camille/Matthias, any of you want to take it up?
ZOOKEEPER-1190: Will take a look at it and will ping Eric for the fix.
ZOOKEEPER-1207: Am very inclined to punt on it. Any takers?
 I am fine with saying RO feature is broken in the release.

I had planned to do an RC this week, but looks like we are close but not there 
yet. Just need a final push. Really appreciate the effort everyone is putting 
to get this through.

thanks
mahadev

PS: Ill be OOO from Friday until Tuesday next week.

Reply via email to