On Tue, Apr 24, 2018 at 3:17 PM, Ravi Shankar Nori <rnori@redhat.com> wrote:
On Tue, Apr 24, 2018 at 7:00 AM, Dan Kenigsberg <danken@redhat.com> wrote:
Ravi's patch is in, but a similar problem remains, and the test cannot be put back into its place.
It seems that while Vdsm was taken down, a couple of getCapsAsync requests queued up. At one point, the host resumed its connection, before the requests have been cleared of the queue. After the host is up, the following tests resume, and at a pseudorandom point in time, an old getCapsAsync request times out and kills our connection.
I believe that as long as ANY request is on flight, the monitoring lock should not be released, and the host should not be declared as up.
Hi Dan,
Can I have the link to the job on jenkins so I can look at the logs
http://jenkins.ovirt.org/job/ovirt-system-tests_standard-check-patch/346/
This [1] should fix the multiple release lock issue
[1] https://gerrit.ovirt.org/#/c/90077/
On Tue, Apr 10, 2018 at 3:53 PM, Ravi Shankar Nori <rnori@redhat.com> wrote:
Working on a patch will post a fix
Thanks
Ravi
On Tue, Apr 10, 2018 at 9:14 AM, Alona Kaplan <alkaplan@redhat.com>
wrote:
Hi all,
Looking at the log it seems that the new GetCapabilitiesAsync is responsible for the mess.
- 08:29:47 - engine loses connectivity to host 'lago-basic-suite-4-2-host-0'.
- Every 3 seconds a getCapabalititiesAsync request is sent to the host (unsuccessfully).
* before each "getCapabilitiesAsync" the monitoring lock is taken (VdsManager,refreshImpl)
* "getCapabilitiesAsync" immediately fails and throws 'VDSNetworkException: java.net.ConnectException: Connection refused'.
The
exception is caught by 'GetCapabilitiesAsyncVDSCommand.executeVdsBrokerCommand' which calls 'onFailure' of the callback and re-throws the exception.
catch (Throwable t) { getParameters().getCallback().onFailure(t); throw t; }
* The 'onFailure' of the callback releases the "monitoringLock" ('postProcessRefresh()->afterRefreshTreatment()-> if (!succeeded) lockManager.releaseLock(monitoringLock);')
* 'VdsManager,refreshImpl' catches the network exception, marks 'releaseLock = true' and tries to release the already released lock.
The following warning is printed to the log -
WARN [org.ovirt.engine.core.bll.lock.InMemoryLockManager] (EE-ManagedThreadFactory-engineScheduled-Thread-53) [] Trying to release exclusive lock which does not exist, lock key: 'ecf53d69-eb68-4b11-8df2-c4aa4e19bd93VDS_INIT'
- 08:30:51 a successful getCapabilitiesAsync is sent.
- 08:32:55 - The failing test starts (Setup Networks for setting ipv6).
* SetupNetworks takes the monitoring lock.
- 08:33:00 - ResponseTracker cleans the getCapabilitiesAsync requests from 4 minutes ago from its queue and prints a VDSNetworkException: Vds timeout occured.
* When the first request is removed from the queue ('ResponseTracker.remove()'), the 'Callback.onFailure' is invoked (for the second time) -> monitoring lock is released (the lock taken by the SetupNetworks!).
* The other requests removed from the queue also try to release
monitoring lock, but there is nothing to release.
* The following warning log is printed - WARN [org.ovirt.engine.core.bll.lock.InMemoryLockManager] (EE-ManagedThreadFactory-engineScheduled-Thread-14) [] Trying to release exclusive lock which does not exist, lock key: 'ecf53d69-eb68-4b11-8df2-c4aa4e19bd93VDS_INIT'
- 08:33:00 - SetupNetwork fails on Timeout ~4 seconds after is started. Why? I'm not 100% sure but I guess the late processing of the 'getCapabilitiesAsync' that causes losing of the monitoring lock and
late + mupltiple processing of failure is root cause.
Ravi, 'getCapabilitiesAsync' failure is treated twice and the lock is trying to be released three times. Please share your opinion regarding how it should be fixed.
Thanks,
Alona.
On Sun, Apr 8, 2018 at 1:21 PM, Dan Kenigsberg <danken@redhat.com> wrote:
On Sun, Apr 8, 2018 at 9:21 AM, Edward Haas <ehaas@redhat.com>
wrote:
> > > > On Sun, Apr 8, 2018 at 9:15 AM, Eyal Edri <eedri@redhat.com> wrote: >> >> Was already done by Yaniv - https://gerrit.ovirt.org/#/c/89851. >> Is it still failing? >> >> On Sun, Apr 8, 2018 at 8:59 AM, Barak Korren <bkorren@redhat.com> >> wrote: >>> >>> On 7 April 2018 at 00:30, Dan Kenigsberg <danken@redhat.com> wrote: >>> > No, I am afraid that we have not managed to understand why setting >>> > and >>> > ipv6 address too the host off the grid. We shall continue >>> > researching >>> > this next week. >>> > >>> > Edy, https://gerrit.ovirt.org/#/c/88637/ is already 4 weeks
On Wed, Apr 11, 2018 at 1:04 AM, Ravi Shankar Nori <rnori@redhat.com> wrote: the the old,
>>> > but >>> > could it possibly be related (I really doubt that)? >>> > > > > Sorry, but I do not see how this problem is related to VDSM. > There is nothing that indicates that there is a VDSM problem. > > Has the RPC connection between Engine and VDSM failed? >
Further up the thread, Piotr noticed that (at least on one failure of this test) that the Vdsm host lost connectivity to its storage, and Vdsm process was restarted. However, this does not seems to happen in all cases where this test fails.
_______________________________________________ Devel mailing list Devel@ovirt.org http://lists.ovirt.org/mailman/listinfo/devel
-- Martin Perina Associate Manager, Software Engineering Red Hat Czech s.r.o.