We suspect that thread.allocate_lock causes issues that leaded to deadlock during long run of vdsm (bz 1022036). As we monkeypatch python's threading package, we should also monkeypatch this lock. The patch format should be: def monkey_patch(): """ Hack threading module to use our classes Thus, Queue and SocketServer can easily enjoy them. """ + import thread + + thread.allocate_lock = Lock + import threading threading.Condition = Condition threading.Lock = Lock threading.RLock = RLock
Hi Yaniv, Please let me know when we have this patch on upstream then I can build a new python-pthreading for fedora too. If we want that to be included on 3.3 probably we will need flags here too.
thread upstream: https://lists.fedorahosted.org/pipermail/vdsm-devel/2014-February/003071.html
The patch posted to https://git.fedorahosted.org/cgit/pthreading.git/ and tar uploaded to https://pypi.python.org/pypi?:action=display&name=pthreading&version=0.1.3
Hello Yaniv, I have build 0.1.3 python-pthreading for Fedora, to have this version of package on RHEV-M-3.4, we need the flags on this bugzilla. Do you want to see this 0.1.3 for EL7 at this moment? python-pthreading-0.1.3-0.fc19 http://koji.fedoraproject.org/koji/buildinfo?buildID=505373 python-pthreading-0.1.3-0.fc20 http://koji.fedoraproject.org/koji/buildinfo?buildID=505371 python-pthreading-0.1.3-0.fc21 http://koji.fedoraproject.org/koji/buildinfo?buildID=505369
all acks granted. yes, el7 build is also required
what to test here?
(In reply to Tareq Alayan from comment #7) > what to test here? Yaniv, can you please share your thoughts here?
same flow you verify in Bug 1022036. read what the customer faces and try to run the system with same load, same duration, until facing the non-responsive state.
since the patch is already riunning on custome env. I am moving it to verified.
Closing as part of 3.4.0