Thu, Apr 18
Wed, Apr 17
Tue, Apr 16
Mon, Apr 15
JFTR; this needs to be dropped from hieradata/common/scap/dsh.yaml
The PHP extension is now also fully deployed (with the exception of the labweb/wikitech hosts, but they still need the migration to PHP 7.2 and are currently HHVM-only)
Fri, Apr 12
We could also look into a backport of https://github.com/systemd/systemd/commit/9009d3b5c3b6d191be69215736be77583e0f23f9 to Stretch, seems totally doable and once confirmed to work fine in our environment, submit it as a merge request for the Debian systemd maintainers for a Stretch point release (every point release ships backports of important bugfixes, e.g. https://tracker.debian.org/news/1037358/accepted-systemd-232-25deb9u10-source-into-proposed-updates-stable-new-proposed-updates/
We need to extend the "raid" fact in modules/raid/lib/facter/raid.rb to also detect the Gen10 controller and then return a custom fact (e.g. "ssa"). modules/raid/manifests/init.pp can then be updated in a subsequent step to automatically install the ssacli tool on the Smart Array Gen10 RAID systems.
HPE renamed the tool, I installed "ssacli" and now "ssacli controller all show config" works fine.
I tried 4.19 on db2102, doesn't make a difference.
Thu, Apr 11
The RAID controller shows up in early device detection by the kernel:
I think we can close this one?
The HHVM extension has been fully rolled out to production. The PHP extension (built from a different source package) is still TBD.
This is weird, do we have a second server of that model for comparison? I don't even see the controller is lspci (it should identify as "Subsystem: Hewlett-Packard Company Smart Array P408i-a SR Gen10"), so I'd like to rule out a hardware/connection issue with that specific server.
These test hosts don't use replication and are standalone, right? I think then we can simply do a "slapcat > foo.ldif" on the old trusty host, then stop slapd on the new stretch host and use slapadd to transfer the LDIF data.
This ticket is superseded by https://phabricator.wikimedia.org/T212772
Wed, Apr 10
He wasn't removed from the cn=wmf LDAP group, I fixed that:
Tue, Apr 9
Mon, Apr 8
During the HHVM updates I noticed that mw2151 is in site.pp as a jobrunner, but not listed in conftool-data.
We currently only have an apt::pin for wikimedia-thumbor, we also need one for librsvg2-2, librsvg2-bin and librsvg2-common.
labtestnet2003 is still in puppetdb:
Fri, Apr 5
resvg is now available in Debian unstable: https://packages.qa.debian.org/r/resvg/news/20190403T150642Z.html
Thu, Apr 4
Is anyone still using Servermon at this point?
Wed, Apr 3
It's a tough nut to crack, I've made progress on a number of issues, but still not fully done yet:
Tue, Apr 2
The later cmake version in combination with the debian/rules file tries to enable position independent ELF files, which doesn't work with libcurl-openssl from standard jessie as it's not yet built as a PIC binary. See the "hardening" section of https://manpages.debian.org/stretch/dpkg-dev/dpkg-buildflags.1.en.html for some background.
I had a look at the missing packages:
Running the steps from the prerm on a jessie system with 8.38 works fine:
Mon, Apr 1
This is done
Fri, Mar 29
Thu, Mar 28
Closing, the remaining work for this is handled via T219333
labtestmetal2001 had /etc/apt/apt.conf.d/00backports-default-release pointing to jessie-backports, which broke debmonitor as jessie-backports has been archived, I removed the file to unbreak it (and given that the host will be rebuilt anyway).
Mar 22 2019
@Legoktm How shall we handle the PHP update of wikidiff2 now that buster is frozen, maybe upload 1.8.1 to experimental and I'll rebuild this in component/php72 fo deployment to production?
I confirm that she has an NDA in place, so I've added uid=julianglen to the nda group in LDAP.
Mar 21 2019
There's already https://phabricator.wikimedia.org/T218544
It could be simply a broken CPU? If we have such the CPU type in a decom host, we could loot it from there.
Mar 18 2019
Mar 15 2019
Fix has been deployed,
Mar 14 2019
Mar 13 2019
Ah, I completely forgot to merge it, it's https://gerrit.wikimedia.org/r/480520, will do that later on
Actually, I had only been looking at servers with php7.2-fpm installed, the deployment, maintenance and snapshot hosts will also need to be converted to the component.
Production has been switched to the new component, all working fine. The new approach is also fairly straightforward, the upgrade from 7.2.15 to 7.2.16 was a straightforward import/build of the new php7.2 source package with all extensions continuing to work fine.
Mar 12 2019
@GTirloni upgraded OpenLDAP on serpens to 2.4.47, but that doesn't change the memory leak.
Mar 11 2019
@MSantos : The service::node class which is used by Kartotherian/Tilerator recently gained an option $use_nodejs10, you can extend the kartotherian/tilerator Puppet classes with a new parameter for node10 which then gets passed down to the service::node class and then enable it in deployment-prep. See modules/aqs/manifests/init.pp which already does that.
The Chromium update has been rolled out, closing the task. I've also notified https://github.com/GoogleChrome/puppeteer/issues/4040 that this seems caused by a Chromium regression.