[ovirt-devel] Re: [VDSM] all test passed, build failed with "tar: journal/b087148aba6d49b9bbef488e52a48752/system.journal: file changed as we read it"

On Sun, Dec 2, 2018 at 8:18 AM Edward Haas <edwardh(a)redhat.com&gt; wrote: > > > On Sat, Dec 1, 2018 at 11:10 PM Nir Soffer <nsoffer(a)redhat.com&gt; wrote: > >> On Thu, Nov 29, 2018 at 11:21 AM Edward Haas <edwardh(a)redhat.com&gt; wrote: >> >>> >>> >>> On Thu, Nov 29, 2018 at 10:41 AM Edward Haas <edwardh(a)redhat.com&gt; >>> wrote: >>> >>>> >>>> >>>> On Wed, Nov 28, 2018 at 8:12 PM Nir Soffer <nsoffer(a)redhat.com&gt; wrote: >>>> >>>>> We have this failure that pops randomly: >>>>> >>>>> 1. All tests pass >>>>> >>>>> *00:13:13.284* ___________________________________ summary ____________________________________*00:13:13.285* tests: commands succeeded*00:13:13.286* storage-py27: commands succeeded*00:13:13.286* storage-py36: commands succeeded*00:13:13.286* lib-py27: commands succeeded*00:13:13.287* lib-py36: commands succeeded*00:13:13.288* network-py27: commands succeeded*00:13:13.290* network-py36: commands succeeded*00:13:13.291* virt-py27: commands succeeded*00:13:13.292* virt-py36: commands succeeded*00:13:13.293* congratulations :) >>>>> >>>>> >>>>> 2. But we fail to collect logs at the end >>>>> >>>>> *00:14:35.992* ##########################################################*00:14:35.995* ## Wed Nov 28 17:39:50 UTC 2018 Finished env: fc28:fedora-28-x86_64*00:14:35.996* ## took 764 seconds*00:14:35.997* ## rc = 1*00:14:35.997* ##########################################################*00:14:36.009* ##! ERROR vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv*00:14:36.010* ##! Last 20 log entries: /tmp/mock_logs.Lcop4ZOq/script/stdout_stderr.log*00:14:36.011* ##!*00:14:36.012* journal/b087148aba6d49b9bbef488e52a48752/system.journal*00:14:36.013* tar: journal/b087148aba6d49b9bbef488e52a48752/system.journal: file changed as we read it*00:14:36.014* journal/b087148aba6d49b9bbef488e52a48752/user-1000.journal*00:14:36.015* lastlog*00:14:36.015* libvirt/*00:14:36.015* libvirt/lxc/*00:14:36.015* libvirt/libxl/*00:14:36.016* libvirt/qemu/*00:14:36.016* libvirt/qemu/LiveOS-f920001d-be4e-47ea-ac26-72480fd5be87.log*00:14:36.017* libvirt/uml/*00:14:36.017* ovirt-guest-agent/*00:14:36.017* ovirt-guest-agent/ovirt-guest-agent.log*00:14:36.017* README*00:14:36.018* samba/*00:14:36.018* samba/old/*00:14:36.018* sssd/*00:14:36.018* tallylog*00:14:36.018* wtmp*00:14:36.018* Took 678 seconds*00:14:36.018* ===================================*00:14:36.019* ##!*00:14:36.019* ##! ERROR ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^*00:14:36.019* ##!######################################################## >>>>> >>>>> >>>>> This looks like an issue with vdsm check-patch.sh: >>>>> >>>>> function collect_logs { >>>>> res=$? >>>>> [ "$res" -ne 0 ] && echo "*** err: $res" >>>>> cd /var/log >>>>> tar -cvzf "$EXPORT_DIR/mock_varlogs.tar.gz" * >>>>> cd /var/host_log >>>>> tar -cvzf "$EXPORT_DIR/host_varlogs.tar.gz" * >>>>> } >>>>> >>>>> trap collect_logs EXIT >>>>> >>>>> Seems that tar fail to collect log if the log is modified while >>>>> copied, which makes sense. >>>>> >>>>> We can ignore errors in tar, since log collection should not fail the >>>>> build, but I think >>>>> a better solution is to avoid collecting any logs since vdsm writes >>>>> its own logs during >>>>> tests - all the info must be in vdsm log. >>>>> >>>>> Here is the list of collected logs: >>>>> >>>>> *00:13:47.280* + tar -cvzf /home/jenkins/workspace/vdsm_master_check-patch-fc28-x86_64/vdsm/exported-artifacts/mock_varlogs.tar.gz btmp dnf.librepo.log dnf.log dnf.rpm.log faillog glusterfs hawkey.log journal lastlog libvirt openvswitch README tallylog vdsm_tests.log wtmp yum.log*00:13:47.285* btmp*00:13:47.285* dnf.librepo.log*00:13:47.299* dnf.log*00:13:47.309* dnf.rpm.log*00:13:47.310* faillog*00:13:47.311* glusterfs/*00:13:47.312* hawkey.log*00:13:47.313* journal/*00:13:47.313* lastlog*00:13:47.315* libvirt/*00:13:47.315* libvirt/qemu/*00:13:47.316* openvswitch/*00:13:47.317* openvswitch/ovs-vswitchd.log*00:13:47.318* openvswitch/ovsdb-server.log*00:13:47.319* README*00:13:47.320* tallylog*00:13:47.321* vdsm_tests.log*00:13:47.342* wtmp*00:13:47.343* yum.log*00:13:47.349* + cd /var/host_log*00:13:47.350* + tar -cvzf /home/jenkins/workspace/vdsm_master_check-patch-fc28-x86_64/vdsm/exported-artifacts/host_varlogs.tar.gz anaconda audit boot.log btmp chrony cloud-init.log cloud-init-output.log cron dnf.librepo.log dnf.log dnf.rpm.log firewalld glusterfs hawkey.log journal lastlog libvirt ovirt-guest-agent README samba sssd tallylog wtmp*00:13:47.356* anaconda/*00:13:47.356* anaconda/ifcfg.log*00:13:47.357* anaconda/ks-script-l5qnynnj.log*00:13:47.358* anaconda/storage.log*00:13:47.359* anaconda/program.log*00:13:47.395* anaconda/ks-script-b5_08tmo.log*00:13:47.396* anaconda/ks-script-6uks8bp3.log*00:13:47.397* anaconda/hawkey.log*00:13:47.398* anaconda/syslog*00:13:47.406* anaconda/journal.log*00:13:47.449* anaconda/dnf.librepo.log*00:13:47.458* anaconda/packaging.log*00:13:47.465* anaconda/dbus.log*00:13:47.466* anaconda/anaconda.log*00:13:47.467* anaconda/ks-script-slrcz39_.log*00:13:47.503* audit/*00:13:47.504* audit/audit.log.3*00:13:47.657* audit/audit.log.2*00:13:47.814* audit/audit.log.1*00:13:47.981* audit/audit.log*00:13:48.008* audit/audit.log.4*00:13:48.155* boot.log*00:13:48.156* btmp*00:13:48.157* chrony/*00:13:48.159* cloud-init.log*00:13:48.159* cloud-init-output.log*00:13:48.161* cron*00:13:48.162* dnf.librepo.log*00:13:49.930* dnf.log*00:13:51.335* dnf.rpm.log*00:13:51.421* firewalld*00:13:51.423* glusterfs/*00:13:51.424* hawkey.log*00:13:51.704* journal/*00:13:51.708* journal/b087148aba6d49b9bbef488e52a48752/*00:13:51.709* journal/b087148aba6d49b9bbef488e52a48752/system.journal*00:13:55.817* tar: journal/b087148aba6d49b9bbef488e52a48752/system.journal: file changed as we read it*00:13:55.819* journal/b087148aba6d49b9bbef488e52a48752/user-1000.journal*00:13:55.915* lastlog*00:13:55.923* libvirt/*00:13:55.924* libvirt/lxc/*00:13:55.926* libvirt/libxl/*00:13:55.927* libvirt/qemu/*00:13:55.928* libvirt/qemu/LiveOS-f920001d-be4e-47ea-ac26-72480fd5be87.log*00:13:55.929* libvirt/uml/*00:13:55.930* ovirt-guest-agent/*00:13:55.930* ovirt-guest-agent/ovirt-guest-agent.log*00:13:55.932* README*00:13:55.933* samba/*00:13:55.933* samba/old/*00:13:55.935* sssd/*00:13:55.935* tallylog*00:13:55.935* wtmp >>>>> >>>>> >>>>> Most if not all are lot relevant to vdsm tests, and should not be >>>>> collected. >>>>> >>>>> This was added in: >>>>> >>>>> commit 9c9c17297433e5a5a49aa19cde10b206e7db61e9 >>>>> Author: Edward Haas <edwardh(a)redhat.com&gt; >>>>> Date: Tue Apr 17 10:53:11 2018 +0300 >>>>> >>>>> automation: Collect logs even when check-patch fails >>>>> >>>>> Change-Id: Idfe07ce6fc55473b1db1d7f16754f559cc5c345a >>>>> Signed-off-by: Edward Haas <edwardh(a)redhat.com&gt; >>>>> >>>>> Reviewed in: >>>>> https://gerrit.ovirt.org/c/90370 >>>>> >>>>> Edward, can you explain why do we need to collect logs during >>>>> check-patch, >>>>> and why do we need to collect all the logs in the system? >>>>> >>>> >>>> check-patch are running unit and integrations tests. >>>> The integration tests are touching the OS and other packages (like >>>> openvswitch). >>>> It was added so we can debug why tests failed. >>>> >>>> I guess we can now separate the unit and integration tests, but it >>>> will not solve >>>> the problem presented here. >>>> Failing to collect the logs silently sounds a good enough solution to >>>> me. >>>> >>> >>> Barak suggested to just exclude the journal: >>> https://gerrit.ovirt.org/#/c/95850/ >>> >> >> This is fixed now, thanks! >> > Not fixed yet, we still fail collecting /var/host_log: + cd /var/host_log + tar --exclude 'journal/*' -czf /home/jenkins/workspace/vdsm_standard-check-patch/vdsm/exported-artifacts/mock_varlogs.tar.gz btmp faillog glusterfs grubby_prune_debug lastlog libvirt openvswitch swtpm tallylog vdsm_tests.log wtmp yum.log + cd /var/host_log + tar -czf /home/jenkins/workspace/vdsm_standard-check-patch/vdsm/exported-artifacts/host_varlogs.tar.gz anaconda audit boot.log boot.log-20181210 boot.log-20181211 boot.log-20181212 boot.log-20181213 boot.log-20181214 boot.log-20181215 boot.log-20181216 btmp btmp-20181201 cron cron-20181126 cron-20181202 cron-20181210 cron-20181216 dmesg dmesg.old firewalld glusterfs grubby grubby_prune_debug httpd journal lastlog libvirt maillog maillog-20181126 maillog-20181202 maillog-20181210 maillog-20181216 messages ntpstats ovirt-engine ovirt-guest-agent ovirt-imageio-proxy ppp puppet qemu-ga secure spooler spooler-20181126 spooler-20181202 spooler-20181210 spooler-20181216 tallylog tuned wpa_supplicant.log wtmp yum.log yum.log-20170101 yum.log-20180101 tar: journal/d2b3276bfc6c7a4e95ce6b2b9b5d0f20/system.journal: file changed as we read it https://jenkins.ovirt.org/blue/rest/organizations/jenkins/pipelines/vdsm_... Copying the binary logs from journal is not the way to copy logs. If we need the host journal, it should be collected by CI infra using journalctl.

Nir > > >> Any reason why we exclude journal only for /var/log, and do collect the >> binary journal >> from /var/host_log? I guess it can fail in the same way. >> > > As far as I know, /var/host_log is collected by the CI and placed (copy) > there. > It does not collect everything, therefore we added the second one. > > >> If we collect logs for integration tests, the most important log is the >> journal, and now >> we skip it. >> > > I think it appears in /var/host_log. > > >> >> We can get the journal in a reliable way like this: >> >> journalctl --since build-date > /tmp/journal.log >> >> Nir >> >>>