ubuntu tripleo-ci jobs are not logstash indexable

Bug #1328645 reported by Robert Collins
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
OpenStack Core Infrastructure
In Progress
Undecided
Patrick Crews

Bug Description

In https://review.openstack.org/#/c/98043 we're going to preserve the hierarchy so that we can have multiple nova-conductor, etc, logs spread out across the various nodes we've test-deployed too, which will permit https://review.openstack.org/#/c/97947/2/toci_devtest.sh to have useful logs for F20 jobs in logstash. YAY.

However we still have the Ubuntu logs all tarred up which stops logstash seeing them.

We can fix this by untarring the logs, but a deep hierarchy will cause infra problems due to poor inode utilisation on the static server(s). This is complicated by us needing two hierarchies - /var/log and /mnt/state/var/log

Discussion with clarkb on IRC, we came up with:

/$instance/root containing /var/log and
/$instance/mnt containing /mnt/state/var/log

Which would avoid concerns on namespace clashes and eliminate 4 pointless inodes per node. We may find there are too many subdirs under /var/log/ etc and need to take further flattening work, but just rsyncing those two paths across (and compressing the contents of the hierarchy e.g. nova-conductor.log -> nova-conductor.log.gz) should give us a good starting point.

Changed in openstack-ci:
assignee: nobody → Patrick Crews (patrick-crews)
Changed in openstack-ci:
status: New → Fix Committed
status: Fix Committed → In Progress
Revision history for this message
Derek Higgins (derekh) wrote :

The log files from the ubuntu tar ball are already being extracted e.g. see
http://logs.openstack.org/95/99495/1/check-tripleo/check-tripleo-ironic-seed-precise/5924db9/logs/seed_logs/

and we're currently stripping out the /var/log/upstart path with tar
--strip-components=3

Revision history for this message
Robert Collins (lifeless) wrote :

so we need to handle /mnt only, cool. I'd still suggest moving to rsync and doing a single-pass operation though, it will be faster than tar + compress + untar.

Revision history for this message
Derek Higgins (derekh) wrote :

Timing for what we currently have tar/gz/scp/untar (picked two overcloud at random one F20 and one Precise)

http://logs.openstack.org/51/93451/3/check-tripleo/check-tripleo-overcloud-f20/2b3bea0/console.html (23 secs)
2014-06-13 06:55:13.602 | Run completed.
2014-06-13 06:55:36.002 | Cleaning up host

http://logs.openstack.org/44/94944/12/experimental-tripleo/check-tripleo-overcloud-precise/0203699/console.html (19 secs)
2014-06-12 17:25:33.183 | Run completed.
2014-06-12 17:25:52.930 | Cleaning up host

1. so we have 2 options, get the /mnt logs from the tarball and nothing else changes
or
2. use rsync

If going down the rsync route before merging can we compare the new times and make sure its faster and simpler

Revision history for this message
Patrick Crews (patrick-crews) wrote :

http://logs.openstack.org/50/99250/6/check-tripleo/check-tripleo-overcloud-f20/805efa1/console.html

rsync appears to be slower:
2014-06-11 08:02:57.514 | Run completed.
2014-06-11 08:03:48.494 | Cleaning up host

Looks to also have some permission errors:
2014-06-11 08:03:30.307 | + rsync -avz -e 'ssh -o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null -o LogLevel=QUIET -o PasswordAuthentication=no' heat-admin@192.0.2.3:/var/log :/etc :/mnt/state/var/log /home/jenkins/workspace/check-tripleo-overcloud-f20/logs/192.0.2.3_logs
2014-06-11 08:03:32.204 | receiving incremental file list
2014-06-11 08:03:32.330 | rsync: opendir "/etc/audisp" failed: Permission denied (13)

Revision history for this message
Patrick Crews (patrick-crews) wrote :

https://review.openstack.org/#/c/99250/ adds /mnt/state/var/log to extracted files.

Also need to update https://github.com/openstack-infra/config/blob/master/modules/openstack_project/files/logstash/jenkins-log-client.yaml to include new files as well as to update (we have old names for over/undercloud nodes)

We may need to change over/under cloud node naming, which currently includes random strings as this does not appear to be terribly friendly to the indexing code.

Revision history for this message
Patrick Crews (patrick-crews) wrote :
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix merged to system-config (master)

Reviewed: https://review.openstack.org/107841
Committed: https://git.openstack.org/cgit/openstack-infra/system-config/commit/?id=f12e712bb922df3a0bbc25a6bbb6cf03ea648127
Submitter: Jenkins
Branch: master

commit f12e712bb922df3a0bbc25a6bbb6cf03ea648127
Author: pcrews <email address hidden>
Date: Thu Jul 17 14:54:38 2014 -0700

    Adding tripleo-ci files to be indexed by logstash

    Adding log files and fixing format tags for tripleo jobs
    in the config file defining which files get indexed by logstash.

    Only adding common / basic files for host_info, os-collect-config,
    and mysql logs from /mnt/state/var/log.

    Partial-bug: 1328645
    Change-Id: I9d858895602440668fbf1f9ecd34bb4e3b9a2548

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.