<div dir="ltr"><div class="gmail_default" style="font-family:arial,helvetica,sans-serif"><br></div><div class="gmail_extra"><br><div class="gmail_quote">On Wed, Aug 30, 2017 at 8:43 PM, Dan Kenigsberg <span dir="ltr"><<a href="mailto:danken@redhat.com" target="_blank">danken@redhat.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">On Wed, Aug 30, 2017 at 6:47 PM, Dominik Holler <<a href="mailto:dholler@redhat.com">dholler@redhat.com</a>> wrote:<br>
> On Wed, 30 Aug 2017 13:31:38 +0200<br>
> Dominik Holler <<a href="mailto:dholler@redhat.com">dholler@redhat.com</a>> wrote:<br>
><br>
>> On Wed, 30 Aug 2017 14:18:49 +0300<br>
>> Dan Kenigsberg <<a href="mailto:danken@redhat.com">danken@redhat.com</a>> wrote:<br>
>><br>
>> > On Wed, Aug 30, 2017 at 1:40 PM, Barak Korren <<a href="mailto:bkorren@redhat.com">bkorren@redhat.com</a>><br>
>> > wrote:<br>
>> > > Test failed: [ 002_bootstrap.add_hosts ]<br>
>> > ><br>
>> > > Link to suspected patches:<br>
>> > ><br>
>> > > We suspect this is due to change to LLDPAD in upstream CentOS<br>
>> > > repos. We can't tell the exact point it was introduced because<br>
>> > > other ovirt-engine regressions introduced too much noise into the<br>
>> > > system.<br>
>> > ><br>
>> > > It also seems that the failure is not 100% reproducible sine we<br>
>> > > have runs that do not encounter it.<br>
>> > ><br>
>> > > Link to Job:<br>
>> > > <a href="http://jenkins.ovirt.org/job/ovirt-master_change-queue-tester/2151" rel="noreferrer" target="_blank">http://jenkins.ovirt.org/job/<wbr>ovirt-master_change-queue-<wbr>tester/2151</a><br>
>> > ><br>
>> > > (This job checked a specific patch to ovirt-log-collector but<br>
>> > > failure seems unrelated, and also happens on vanille 'tested' repo<br>
>> > > right now).<br>
>> > ><br>
>> > > Link to all logs:<br>
>> > > VDSM logs seem to be most relevant here:<br>
>> > ><br>
>> > > <a href="http://jenkins.ovirt.org/job/ovirt-master_change-queue-tester/2151/artifact/exported-artifacts/basic-suit-master-el7/test_logs/basic-suite-master/post-002_bootstrap.py/lago-basic-suite-master-host-0/_var_log/vdsm/" rel="noreferrer" target="_blank">http://jenkins.ovirt.org/job/<wbr>ovirt-master_change-queue-<wbr>tester/2151/artifact/exported-<wbr>artifacts/basic-suit-master-<wbr>el7/test_logs/basic-suite-<wbr>master/post-002_bootstrap.py/<wbr>lago-basic-suite-master-host-<wbr>0/_var_log/vdsm/</a><br>
>> > ><br>
>> > > Error snippet from log:<br>
>> > ><br>
>> > > - From VDSM logs (note it exists very quickly):<br>
>> > ><br>
>> > > <error><br>
>> > ><br>
>> > > 2017-08-30 05:55:52,208-0400 INFO (MainThread) [vds] (PID: 4594)<br>
>> > > I am the actual vdsm 4.20.2-133.git0ce4485.el7.<wbr>centos<br>
>> > > lago-basic-suite-master-host-0 (3.10.0-514.2.2.el7.x86_64)<br>
>> > > (vdsmd:148) 2017-08-30 05:55:52,209-0400 INFO (MainThread) [vds]<br>
>> > > VDSM will run with cpu affinity: frozenset([1]) (vdsmd:254)<br>
>> > > 2017-08-30 05:55:52,253-0400 INFO (MainThread) [storage.check]<br>
>> > > Starting check service (check:92)<br>
>> > > 2017-08-30 05:55:52,257-0400 INFO (MainThread)<br>
>> > > [storage.Dispatcher] Starting StorageDispatcher... (dispatcher:48)<br>
>> > > 2017-08-30 05:55:52,257-0400 INFO (check/loop)<br>
>> > > [storage.asyncevent] Starting <EventLoop running=True<br>
>> > > closed=False at 0x45317520> (asyncevent:125)<br>
>> > > 2017-08-30 05:55:52,288-0400 INFO (MainThread) [vdsm.api] START<br>
>> > > registerDomainStateChangeCallb<wbr>ack(callbackFunc=<functools.<wbr>partial<br>
>> > > object at 0x2b47b50>) from=internal,<br>
>> > > task_id=2cebe9ef-358e-47d7-<wbr>81a6-8c4b54b9cd6d (api:46)<br>
>> > > 2017-08-30 05:55:52,288-0400 INFO (MainThread) [vdsm.api] FINISH<br>
>> > > registerDomainStateChangeCallb<wbr>ack return=None from=internal,<br>
>> > > task_id=2cebe9ef-358e-47d7-<wbr>81a6-8c4b54b9cd6d (api:52)<br>
>> > > 2017-08-30 05:55:52,288-0400 INFO (MainThread) [MOM] Preparing<br>
>> > > MOM interface (momIF:53)<br>
>> > > 2017-08-30 05:55:52,289-0400 INFO (MainThread) [MOM] Using named<br>
>> > > unix socket /var/run/vdsm/mom-vdsm.sock (momIF:62)<br>
>> > > 2017-08-30 05:55:52,289-0400 INFO (MainThread) [root]<br>
>> > > Unregistering all secrets (secret:91)<br>
>> > > 2017-08-30 05:55:52,307-0400 INFO (MainThread) [vds] Setting<br>
>> > > channels' timeout to 30 seconds. (vmchannels:221)<br>
>> > > 2017-08-30 05:55:52,309-0400 INFO (vmrecovery) [vds] recovery:<br>
>> > > completed in 0s (clientIF:516)<br>
>> > > 2017-08-30 05:55:52,310-0400 INFO (MainThread)<br>
>> > > [vds.MultiProtocolAcceptor] Listening at :::54321<br>
>> > > (protocoldetector:196)<br>
>> > > 2017-08-30 05:55:52,496-0400 INFO (http) [vds] Server running<br>
>> > > (http:58) 2017-08-30 05:55:52,742-0400 INFO (periodic/1)<br>
>> > > [vdsm.api] START repoStats(domains=()) from=internal,<br>
>> > > task_id=85768015-8ecb-48e3-<wbr>9307-f671bfc33c65 (api:46)<br>
>> > > 2017-08-30 05:55:52,743-0400 INFO (periodic/1) [vdsm.api] FINISH<br>
>> > > repoStats return={} from=internal,<br>
>> > > task_id=85768015-8ecb-48e3-<wbr>9307-f671bfc33c65 (api:52)<br>
>> > > 2017-08-30 05:55:52,743-0400 WARN (periodic/1) [throttled] MOM<br>
>> > > not available. (throttledlog:103)<br>
>> > > 2017-08-30 05:55:52,744-0400 WARN (periodic/1) [throttled] MOM<br>
>> > > not available, KSM stats will be missing. (throttledlog:103)<br>
>> > > 2017-08-30 05:55:55,043-0400 INFO (MainThread) [vds] Received<br>
>> > > signal 15, shutting down (vdsmd:67)<br>
>> > > 2017-08-30 05:55:55,045-0400 INFO (MainThread)<br>
>> > > [jsonrpc.JsonRpcServer] Stopping JsonRPC Server (__init__:759)<br>
>> > > 2017-08-30 05:55:55,049-0400 INFO (MainThread) [vds] Stopping<br>
>> > > http server (http:79)<br>
>> > > 2017-08-30 05:55:55,049-0400 INFO (http) [vds] Server stopped<br>
>> > > (http:69) 2017-08-30 05:55:55,050-0400 INFO (MainThread) [root]<br>
>> > > Unregistering all secrets (secret:91)<br>
>> > > 2017-08-30 05:55:55,052-0400 INFO (MainThread) [vdsm.api] START<br>
>> > > prepareForShutdown(options=<wbr>None) from=internal,<br>
>> > > task_id=ffde5caa-fa44-49ab-<wbr>bdd1-df81519680a3 (api:46)<br>
>> > > 2017-08-30 05:55:55,089-0400 INFO (MainThread) [storage.Monitor]<br>
>> > > Shutting down domain monitors (monitor:222)<br>
>> > > 2017-08-30 05:55:55,090-0400 INFO (MainThread) [storage.check]<br>
>> > > Stopping check service (check:105)<br>
>> > > 2017-08-30 05:55:55,090-0400 INFO (check/loop)<br>
>> > > [storage.asyncevent] Stopping <EventLoop running=True<br>
>> > > closed=False at 0x45317520> (asyncevent:220)<br>
>> > > 2017-08-30 05:55:55,090-0400 INFO (MainThread) [vdsm.api] FINISH<br>
>> > > prepareForShutdown return=None from=internal,<br>
>> > > task_id=ffde5caa-fa44-49ab-<wbr>bdd1-df81519680a3 (api:52)<br>
>> > > 2017-08-30 05:55:55,091-0400 INFO (MainThread) [vds] Stopping<br>
>> > > threads (vdsmd:159)<br>
>> > > 2017-08-30 05:55:55,091-0400 INFO (MainThread) [vds] Exiting<br>
>> > > (vdsmd:170)<br>
>> > ><br>
>> > > </error><br>
>> > ><br>
>> > ><br>
>> > > - From SuperVDSM logs:<br>
>> > ><br>
>> > > <error><br>
>> > ><br>
>> > > MainThread::ERROR::2017-08-30<br>
>> > > 05:55:59,476::initializer::53:<wbr>:root::(_lldp_init) Failed to enable<br>
>> > > LLDP on eth1<br>
>> > > Traceback (most recent call last):<br>
>> > > File<br>
>> > > "/usr/lib/python2.7/site-<wbr>packages/vdsm/network/<wbr>initializer.py",<br>
>> > > line 51, in _lldp_init Lldp.enable_lldp_on_iface(<wbr>device)<br>
>> > > File<br>
>> > > "/usr/lib/python2.7/site-<wbr>packages/vdsm/network/lldp/<wbr>lldpad.py",<br>
>> > > line 30, in enable_lldp_on_iface<br>
>> > > lldptool.enable_lldp_on_iface(<wbr>iface, rx_only) File<br>
>> > > "/usr/lib/python2.7/site-<wbr>packages/vdsm/network/lldpad/<wbr>lldptool.py",<br>
>> > > line 46, in enable_lldp_on_iface raise EnableLldpError(rc, out,<br>
>> > > err, iface) EnableLldpError: (1,<br>
>> > > "timeout\n'<wbr>M00000001C3040000000c04eth1000<wbr>badminStatus0002rx'<br>
>> > > command timed out.\n", '', 'eth1')<br>
>> > > MainThread::DEBUG::2017-08-30<br>
>> > > 05:55:59,477::cmdutils::133::<wbr>root::(exec_cmd) /usr/sbin/lldptool<br>
>> > > get-lldp -i eth0 adminStatus (cwd None)<br>
>> > ><br>
>> > ><br>
>> > > </error><br>
>> ><br>
>> > The LLDP errors are due to SELinux, to be fixed in CentOS-7.4.1.<br>
>> > They do not cause supervdsmd to stop, so they are not the reason<br>
>> > for vdsm's failure to start. Something ELSE is doing this.<br>
>> ><br>
>> > in<br>
>> > <a href="http://jenkins.ovirt.org/job/ovirt-master_change-queue-tester/2151/artifact/exported-artifacts/basic-suit-master-el7/test_logs/basic-suite-master/post-002_bootstrap.py/lago-basic-suite-master-host-0/_var_log/messages/*view*/" rel="noreferrer" target="_blank">http://jenkins.ovirt.org/job/<wbr>ovirt-master_change-queue-<wbr>tester/2151/artifact/exported-<wbr>artifacts/basic-suit-master-<wbr>el7/test_logs/basic-suite-<wbr>master/post-002_bootstrap.py/<wbr>lago-basic-suite-master-host-<wbr>0/_var_log/messages/*view*/</a><br>
>> > we have<br>
>> ><br>
>> > Aug 30 05:55:52 lago-basic-suite-master-host-0 journal: vdsm<br>
>> > throttled WARN MOM not available.<br>
>> > Aug 30 05:55:52 lago-basic-suite-master-host-0 journal: vdsm<br>
>> > throttled WARN MOM not available, KSM stats will be missing.<br>
>> > Aug 30 05:55:53 lago-basic-suite-master-host-0 systemd: Started<br>
>> > Session 10 of user root.<br>
>> > Aug 30 05:55:53 lago-basic-suite-master-host-0 systemd-logind: New<br>
>> > session 10 of user root.<br>
>> > Aug 30 05:55:53 lago-basic-suite-master-host-0 systemd: Starting<br>
>> > Session 10 of user root.<br>
>> > Aug 30 05:55:53 lago-basic-suite-master-host-0 python: ansible-setup<br>
>> > Invoked with filter=* gather_subset=['all']<br>
>> > fact_path=/etc/ansible/facts.d gather_timeout=10<br>
>> > Aug 30 05:55:54 lago-basic-suite-master-host-0 python:<br>
>> > ansible-command Invoked with warn=True executable=None<br>
>> > _uses_shell=False _raw_params=bash -c "rpm -qi vdsm | grep -oE<br>
>> > 'Version\\s+:\\s+[0-9\\.]+' | awk '{print $3}'" removes=None<br>
>> > creates=None chdir=None<br>
>> > Aug 30 05:55:54 lago-basic-suite-master-host-0 python:<br>
>> > ansible-systemd Invoked with no_block=False name=libvirt-guests<br>
>> > enabled=True daemon_reload=False state=started user=False<br>
>> > masked=None Aug 30 05:55:54 lago-basic-suite-master-host-0 systemd:<br>
>> > Reloading. Aug 30 05:55:55 lago-basic-suite-master-host-0 systemd:<br>
>> > Cannot add dependency job for unit lvm2-lvmetad.socket, ignoring:<br>
>> > Unit is masked. Aug 30 05:55:55 lago-basic-suite-master-host-0<br>
>> > systemd: Stopped MOM instance configured for VDSM purposes.<br>
>> > Aug 30 05:55:55 lago-basic-suite-master-host-0 systemd: Stopping<br>
>> > Virtual Desktop Server Manager...<br>
>> > Aug 30 05:55:55 lago-basic-suite-master-host-0 systemd: Starting<br>
>> > Suspend Active Libvirt Guests...<br>
>> > Aug 30 05:55:55 lago-basic-suite-master-host-0 systemd: Started<br>
>> > Suspend Active Libvirt Guests.<br>
>> > Aug 30 05:55:55 lago-basic-suite-master-host-0 journal: libvirt<br>
>> > version: 2.0.0, package: 10.el7_3.9 (CentOS BuildSystem<br>
>> > <<a href="http://bugs.centos.org" rel="noreferrer" target="_blank">http://bugs.centos.org</a>>, 2017-05-25-20:52:28, <a href="http://c1bm.rdu2.centos.org" rel="noreferrer" target="_blank">c1bm.rdu2.centos.org</a>)<br>
>> > Aug 30 05:55:55 lago-basic-suite-master-host-0 journal: hostname:<br>
>> > lago-basic-suite-master-host-<wbr>0.lago.local<br>
>> > Aug 30 05:55:55 lago-basic-suite-master-host-0 vdsmd_init_common.sh:<br>
>> > vdsm: Running run_final_hooks<br>
>> > Aug 30 05:55:55 lago-basic-suite-master-host-0 journal: End of file<br>
>> > while reading data: Input/output error<br>
>> > Aug 30 05:55:55 lago-basic-suite-master-host-0 systemd: Stopped<br>
>> > Virtual Desktop Server Manager.<br>
>> ><br>
>> > suggesting that systemd stopped vdsmd since mom did not start on<br>
>> > time. We have an extremely ugly two-way dependency between the two<br>
>> > services. Each depends on the other one. I am guessing that<br>
>> > systemd's grace period for such a cycle has expired, and the couple<br>
>> > was taken down.<br>
>> ><br>
>> > Does anybody have a guess what could cause such slowlness?<br>
>><br>
>> Even if the LLDP errors does not have a negative impact of the<br>
>> functional behavior of supervdsmd, they have a impact timing behavior<br>
>> of the initialization of supervdsmd.<br>
>><br>
>> I will post a patch today which moves the initialization of LLDP in an<br>
>> extra thread, to make the initialization of LLDP asynchronly.<br>
>><br>
>><br>
><br>
> I analyzed this issue further by running OST [1] while disabled [2] the<br>
> suspected initialization of LLDP in supervdsmd.<br>
> Like expected, there are no log entries about enabling LLDP in [3],<br>
> but the OST is still failing with same symptoms.<br>
><br>
> So the reason for this failure is still unknown.<br>
><br>
><br>
> [1]<br>
> <a href="http://jenkins.ovirt.org/view/oVirt%20system%20tests/job/ovirt-system-tests_manual/1045/" rel="noreferrer" target="_blank">http://jenkins.ovirt.org/view/<wbr>oVirt%20system%20tests/job/<wbr>ovirt-system-tests_manual/<wbr>1045/</a><br>
><br>
> [2]<br>
> <a href="https://gerrit.ovirt.org/#/c/81232/1/lib/vdsm/network/initializer.py" rel="noreferrer" target="_blank">https://gerrit.ovirt.org/#/c/<wbr>81232/1/lib/vdsm/network/<wbr>initializer.py</a><br>
><br>
> [3]<br>
> <a href="http://jenkins.ovirt.org/view/oVirt%20system%20tests/job/ovirt-system-tests_manual/1045/artifact/exported-artifacts/test_logs/basic-suite-master/post-002_bootstrap.py/lago-basic-suite-master-host-0/_var_log/vdsm/supervdsm.log/*view*/" rel="noreferrer" target="_blank">http://jenkins.ovirt.org/view/<wbr>oVirt%20system%20tests/job/<wbr>ovirt-system-tests_manual/<wbr>1045/artifact/exported-<wbr>artifacts/test_logs/basic-<wbr>suite-master/post-002_<wbr>bootstrap.py/lago-basic-suite-<wbr>master-host-0/_var_log/vdsm/<wbr>supervdsm.log/*view*/</a><br>
><br>
<br>
So we're back in square one.<br>
Another possible culprit may be ansible: Vdsm is stopped two seconds<br>
after it logs to the host.<br>
<br>
Aug 30 11:26:24 lago-basic-suite-master-host-0 systemd: Starting<br>
Session 10 of user root.<br>
Aug 30 11:26:25 lago-basic-suite-master-host-0 python: ansible-setup<br>
Invoked with filter=* gather_subset=['all']<br>
fact_path=/etc/ansible/facts.d gather_timeout=10<br>
Aug 30 11:26:25 lago-basic-suite-master-host-0 python: ansible-command<br>
Invoked with warn=True executable=None _uses_shell=False<br>
_raw_params=bash -c "rpm -qi vdsm | grep -oE<br>
'Version\\s+:\\s+[0-9\\.]+' | awk '{print $3}'" removes=None<br>
creates=None chdir=None<br>
Aug 30 11:26:26 lago-basic-suite-master-host-0 python: ansible-systemd<br>
Invoked with no_block=False name=libvirt-guests enabled=True<br>
daemon_reload=False state=started user=False masked=None<br>
Aug 30 11:26:26 lago-basic-suite-master-host-0 systemd: Reloading.<br>
Aug 30 11:26:26 lago-basic-suite-master-host-0 systemd: Cannot add<br>
dependency job for unit lvm2-lvmetad.socket, ignoring: Unit is masked.<br>
Aug 30 11:26:26 lago-basic-suite-master-host-0 systemd: Stopped MOM<br>
instance configured for VDSM purposes.<br>
Aug 30 11:26:26 lago-basic-suite-master-host-0 systemd: Stopping<br>
Virtual Desktop Server Manager...<br>
<br>
<br>
could it be that it triggers a systemd-reload that makes systemd croak<br>
on the vdsm-mom cycle?<br></blockquote><div><br><div style="font-family:arial,helvetica,sans-serif;display:inline" class="gmail_default">We are not restarting VDSM within ovirt-host-deploy Ansible role, the VDSM restart is performed in host-deploy part same as in previous versions.</div> <div style="font-family:arial,helvetica,sans-serif;display:inline" class="gmail_default">Within ovirt-host-deploy-firewalld we only enable and restart firewalld service.<br></div></div></div><br></div></div>