Right now the MediaWiki deploy servers (e.g., deploy1002) are included in /etc/dsh/group/mediawiki-installation. This means deploy servers have both /srv/mediawiki-staging and /srv/mediawiki directories. This can result in confusion of the type described in T253547.
Proposal:
| Title | Reference | Author | Source Branch | Dest Branch | |
|---|---|---|---|---|---|
| Revert "Exclude deploy servers from target list if /srv/mediawiki is a symlink" | repos/releng/scap!128 | dancy | review/dancy/T329857-cleanup | master | |
| Exclude deploy servers from target list if /srv/mediawiki is a symlink | repos/releng/scap!103 | dancy | T329857 | master | |
| Make scap pull exit if /srv/mediawiki is a symlink | repos/releng/scap!91 | dancy | review/dancy/no-pull-via-symlink | master |
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | Krinkle | T253547 Command line profiling not working on production | |||
| Resolved | dancy | T329857 MediaWiki deploy servers should not be mediawiki installation targets |
dancy opened https://gitlab.wikimedia.org/repos/releng/scap/-/merge_requests/91
Make scap pull exit if /srv/mediawiki is a symlink
dancy merged https://gitlab.wikimedia.org/repos/releng/scap/-/merge_requests/91
Make scap pull exit if /srv/mediawiki is a symlink
Mentioned in SAL (#wikimedia-operations) [2023-03-13T15:21:19Z] <dancy@deploy2002> Started scap: testing T329857
Mentioned in SAL (#wikimedia-operations) [2023-03-13T15:31:28Z] <dancy@deploy2002> Finished scap: testing T329857 (duration: 10m 08s)
dancy opened https://gitlab.wikimedia.org/repos/releng/scap/-/merge_requests/103
Exclude deploy servers from target list if /srv/mediawiki is a symlink
jnuche merged https://gitlab.wikimedia.org/repos/releng/scap/-/merge_requests/103
Exclude deploy servers from target list if /srv/mediawiki is a symlink
Mentioned in SAL (#wikimedia-operations) [2023-03-13T17:15:27Z] <dancy@deploy2002> Started scap: testing T329857
Mentioned in SAL (#wikimedia-operations) [2023-03-13T17:22:22Z] <dancy@deploy2002> Finished scap: testing T329857 (duration: 06m 54s)
Change 901667 had a related patch set uploaded (by Ahmon Dancy; author: Ahmon Dancy):
[operations/puppet@production] Add setting to make /srv/mediawiki -> /srv/mediawiki-staging on deploy servers
Change 901676 had a related patch set uploaded (by Ahmon Dancy; author: Ahmon Dancy):
[operations/puppet@production] Experiment: Remove deploy1002/deploy2002 from mediawiki-installation dsh group
Change 901667 merged by Clément Goubert:
[operations/puppet@production] Add setting to make /srv/mediawiki -> /srv/mediawiki-staging on deploy servers
Mentioned in SAL (#wikimedia-operations) [2023-04-03T13:50:54Z] <claime> Testing deploy server dsh group inclusion - T329857
Change 901676 merged by Clément Goubert:
[operations/puppet@production] Experiment: Remove deploy1002/deploy2002 from mediawiki-installation dsh group
deploy2002:
Notice: /Stage[main]/Scap::Dsh/Scap::Dsh::Group[mediawiki-installation]/Confd::File[/etc/dsh/group/mediawiki-installation]/File[/etc/confd/templates/_etc_dsh_group_mediawiki-installation.tmpl]/content: --- /etc/confd/templates/_etc_dsh_group_mediawiki-installation.tmpl 2023-03-20 20:29:22.934389074 +0000 +++ /tmp/puppet-file20230403-32678-1okp5h1 2023-04-03 13:53:13.660934333 +0000 @@ -7,8 +7,6 @@ cloudweb1003.wikimedia.org cloudweb1004.wikimedia.org cloudweb2002-dev.wikimedia.org -deploy1002.eqiad.wmnet -deploy2002.codfw.wmnet mwmaint1002.eqiad.wmnet mwmaint2002.codfw.wmnet scandium.eqiad.wmnet Info: Computing checksum on file /etc/confd/templates/_etc_dsh_group_mediawiki-installation.tmpl Info: /Stage[main]/Scap::Dsh/Scap::Dsh::Group[mediawiki-installation]/Confd::File[/etc/dsh/group/mediawiki-installation]/File[/etc/confd/templates/_etc_dsh_group_mediawiki-installation.tmpl]: Filebucketed /etc/confd/templates/_etc_dsh_group_mediawiki-installation.tmpl to puppet with sum ab57d93fdbf1598fb7078912cf2b272a Notice: /Stage[main]/Scap::Dsh/Scap::Dsh::Group[mediawiki-installation]/Confd::File[/etc/dsh/group/mediawiki-installation]/File[/etc/confd/templates/_etc_dsh_group_mediawiki-installation.tmpl]/content: content changed '{md5}ab57d93fdbf1598fb7078912cf2b272a' to '{md5}4922782aad3f36dc2d0f2fe62372e2a3' Info: Confd::File[/etc/dsh/group/mediawiki-installation]: Scheduling refresh of Service[confd] Notice: /Stage[main]/Confd/Base::Service_unit[confd]/Service[confd]: Triggered 'refresh' from 1 event Notice: Applied catalog in 40.76 seconds cgoubert@deploy2002:~$ grep deploy /etc/dsh/group/mediawiki-installation
deploy1002:
cgoubert@deploy1002:~$ grep deploy /etc/dsh/group/mediawiki-installation
Test done, patch reverted.
deploy2002:
Notice: /Stage[main]/Scap::Dsh/Scap::Dsh::Group[mediawiki-installation]/Confd::File[/etc/dsh/group/mediawiki-installation]/File[/etc/confd/templates/_etc_ds h_group_mediawiki-installation.tmpl]/content: --- /etc/confd/templates/_etc_dsh_group_mediawiki-installation.tmpl 2023-04-03 13:53:13.700934508 +0000 +++ /tmp/puppet-file20230403-5284-14h3zc8 2023-04-03 13:58:12.614235535 +0000 @@ -7,6 +7,8 @@ cloudweb1003.wikimedia.org cloudweb1004.wikimedia.org cloudweb2002-dev.wikimedia.org +deploy1002.eqiad.wmnet +deploy2002.codfw.wmnet mwmaint1002.eqiad.wmnet mwmaint2002.codfw.wmnet scandium.eqiad.wmnet Info: Computing checksum on file /etc/confd/templates/_etc_dsh_group_mediawiki-installation.tmpl Info: /Stage[main]/Scap::Dsh/Scap::Dsh::Group[mediawiki-installation]/Confd::File[/etc/dsh/group/mediawiki-installation]/File[/etc/confd/templates/_etc_dsh_ group_mediawiki-installation.tmpl]: Filebucketed /etc/confd/templates/_etc_dsh_group_mediawiki-installation.tmpl to puppet with sum 4922782aad3f36dc2d0f2fe6 2372e2a3 Notice: /Stage[main]/Scap::Dsh/Scap::Dsh::Group[mediawiki-installation]/Confd::File[/etc/dsh/group/mediawiki-installation]/File[/etc/confd/templates/_etc_ds h_group_mediawiki-installation.tmpl]/content: content changed '{md5}4922782aad3f36dc2d0f2fe62372e2a3' to '{md5}ab57d93fdbf1598fb7078912cf2b272a' Info: Confd::File[/etc/dsh/group/mediawiki-installation]: Scheduling refresh of Service[confd] Notice: /Stage[main]/Confd/Base::Service_unit[confd]/Service[confd]: Triggered 'refresh' from 1 event Notice: Applied catalog in 42.01 seconds cgoubert@deploy2002:~$ grep deploy /etc/dsh/group/mediawiki-installation deploy1002.eqiad.wmnet deploy2002.codfw.wmnet
deploy1002:
Notice: /Stage[main]/Scap::Dsh/Scap::Dsh::Group[mediawiki-installation]/Confd::File[/etc/dsh/group/mediawiki-installation]/File[/etc/confd/templates/_etc_dsh_group_mediawiki-installation.tmpl]/content: --- /etc/confd/templates/_etc_dsh_group_mediawiki-installation.tmpl 2023-04-03 13:53:43.292813651 +0000 +++ /tmp/puppet-file20230403-17648-15z6t1l 2023-04-03 13:58:27.258069371 +0000 @@ -7,6 +7,8 @@ cloudweb1003.wikimedia.org cloudweb1004.wikimedia.org cloudweb2002-dev.wikimedia.org +deploy1002.eqiad.wmnet +deploy2002.codfw.wmnet mwmaint1002.eqiad.wmnet mwmaint2002.codfw.wmnet scandium.eqiad.wmnet Info: Computing checksum on file /etc/confd/templates/_etc_dsh_group_mediawiki-installation.tmpl Info: /Stage[main]/Scap::Dsh/Scap::Dsh::Group[mediawiki-installation]/Confd::File[/etc/dsh/group/mediawiki-installation]/File[/etc/confd/templates/_etc_dsh_group_mediawiki-installation.tmpl]: Filebucketed /etc/confd/templates/_etc_dsh_group_mediawiki-installation.tmpl to puppet with sum 4922782aad3f36dc2d0f2fe62372e2a3 Notice: /Stage[main]/Scap::Dsh/Scap::Dsh::Group[mediawiki-installation]/Confd::File[/etc/dsh/group/mediawiki-installation]/File[/etc/confd/templates/_etc_dsh_group_mediawiki-installation.tmpl]/content: content changed '{md5}4922782aad3f36dc2d0f2fe62372e2a3' to '{md5}ab57d93fdbf1598fb7078912cf2b272a' Info: Confd::File[/etc/dsh/group/mediawiki-installation]: Scheduling refresh of Service[confd] Notice: /Stage[main]/Confd/Base::Service_unit[confd]/Service[confd]: Triggered 'refresh' from 1 event Notice: Applied catalog in 46.29 seconds cgoubert@deploy1002:~$ grep deploy /etc/dsh/group/mediawiki-installation deploy1002.eqiad.wmnet deploy2002.codfw.wmnet
Change 905297 had a related patch set uploaded (by Ahmon Dancy; author: Ahmon Dancy):
[operations/puppet@production] beta: Enable /srv/mediawiki symlink on deployment-deploy03
Change 905304 had a related patch set uploaded (by Ahmon Dancy; author: Ahmon Dancy):
[operations/puppet@production] mediawiki::scap: Ensure Exec['fetch_mediawiki'] resource always exists
Change 905304 merged by Clément Goubert:
[operations/puppet@production] mediawiki::scap: Ensure Exec['fetch_mediawiki'] resource always exists
Change 905297 merged by Clément Goubert:
[operations/puppet@production] beta: Enable /srv/mediawiki symlink on deployment-deploy03
Change 906051 had a related patch set uploaded (by Ahmon Dancy; author: Ahmon Dancy):
[operations/puppet@production] mediawiki::scap: force creation of the symlink when enabled
Change 906051 merged by Clément Goubert:
[operations/puppet@production] mediawiki::scap: force creation of the symlink when enabled
Change 905983 had a related patch set uploaded (by Ahmon Dancy; author: Ahmon Dancy):
[operations/puppet@production] Revert "mediawiki::scap: force creation of the symlink when enabled"
Change 905983 merged by Clément Goubert:
[operations/puppet@production] Revert "mediawiki::scap: force creation of the symlink when enabled"
@dancy I've tested this in the Beta Cluster as follows, on the deployment-deploy03 host:
krinkle@deployment-deploy03:~$ PHP='php -d auto_prepend_file=/srv/mediawiki/wmf-config/PhpAutoPrepend.php' mwscript showJobs.php --wiki testwiki --profiler text 0 <!-- 100.00% 130.019 1 - main() .. 5.72% 7.439 1 - ShowJobs::execute 2.20% 2.856 1 - Maintenance::shutdown 0.80% 1.039 99 - JobQueue::__construct -->
And it 1) doesn't crash, and 2) given this host doubles as both mwdeploy and mwdebug role also produces profiling output. The same in production, currently crashes on deploy2002 as per T253547#8576045. After this change, I expect that it will not crash in production and (only mwdebug) produce profiling output.
For context, the -d parameter is a php.ini override that we're looking to puppetize by default on all MW servers (and indirectly also on the deploy host). Hence it's important that it not crash, but only needs to produce profiling output on mwdebug where php-tideways is installed. We already puppetize this INI setting for all web-based php-fpm contexts. What we're adding is puppetizing it for php-cli as well. Hence it will affect the deployment server and Scap's use of maintenance scripts.
Long story short: LGTM, achieves the intended effect from my POV!
Change 909302 had a related patch set uploaded (by Ahmon Dancy; author: Ahmon Dancy):
[operations/puppet@production] Enable /srv/mediawiki symlink on prod deploy servers
Mentioned in SAL (#wikimedia-operations) [2023-04-24T08:09:34Z] <claime> Deploying 909302 on deploy1002 for T329857
Mentioned in SAL (#wikimedia-operations) [2023-04-24T08:10:29Z] <claime> Disabling puppet on deploy2002 - T329857
Change 909302 merged by Clément Goubert:
[operations/puppet@production] Enable /srv/mediawiki symlink on prod deploy servers
Mentioned in SAL (#wikimedia-operations) [2023-04-24T08:14:23Z] <claime> Deploying 909302 on deploy2002 for T329857
cgoubert@deploy1002:~$ PHP='php -d auto_prepend_file=/srv/mediawiki/wmf-config/PhpAutoPrepend.php' \ > mwscript showJobs.php --wiki testwiki --profiler text 0 cgoubert@deploy1002:~$ ls -l /srv/mediawiki lrwxrwxrwx 1 root root 22 Apr 24 08:13 /srv/mediawiki -> /srv/mediawiki-staging
cgoubert@deploy2002:~$ PHP='php -d auto_prepend_file=/srv/mediawiki/wmf-config/PhpAutoPrepend.php' \ > mwscript showJobs.php --wiki testwiki --profiler text 0 cgoubert@deploy2002:~$ ls -l /srv/mediawiki lrwxrwxrwx 1 root root 22 Apr 24 08:15 /srv/mediawiki -> /srv/mediawiki-staging
Mentioned in SAL (#wikimedia-operations) [2023-04-24T08:18:16Z] <cgoubert@deploy2002> Started scap: testing T329857
Mentioned in SAL (#wikimedia-operations) [2023-04-24T08:32:45Z] <cgoubert@deploy2002> Finished scap: testing T329857 (duration: 14m 29s)
Everything looks good from my side, tell me when you have a chance to check and I can remove the backup dirs.
Change 911259 had a related patch set uploaded (by Clément Goubert; author: Clément Goubert):
[operations/puppet@production] P:mediawiki::common: Remove deploy check_dsh_group
Change 911259 merged by Clément Goubert:
[operations/puppet@production] P:mediawiki::common: Remove deploy check_dsh_group
dancy opened https://gitlab.wikimedia.org/repos/releng/scap/-/merge_requests/128
Revert "Exclude deploy servers from target list if /srv/mediawiki is a symlink"
dancy merged https://gitlab.wikimedia.org/repos/releng/scap/-/merge_requests/128
Revert "Exclude deploy servers from target list if /srv/mediawiki is a symlink"
@Clement_Goubert I noticed the /srv/mediawiki.old.20230424.T329857 directory on deploy1002.eqiad.wmnet today. It's safe to delete.