It's currently still running PHP. Should be running HHVM.
Description
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Resolved | Legoktm | T75901 Drop PHP 5.3 support | |||
Declined | • demon | T91590 [Spike] Try out hack (<?hh) for mediawiki-config | |||
Resolved | Joe | T104147 can we get rid of rsvg security patch? | |||
Resolved | Reedy | T94149 Get rid of Zend 5.5 tests for wmf branches | |||
Resolved | None | T86081 Complete the use of HHVM over Zend PHP on the Wikimedia cluster | |||
Resolved | bd808 | T166396 Program 1 Outcome 4: VPS hosting | |||
Resolved | Jdforrester-WMF | T172165 Require either PHP 7.0+ or HHVM in MW 1.31 | |||
Resolved | RobH | T168559 decom silver (was silver has trouble rebooting) | |||
Resolved | Jclark-ctr | T189921 decom californium | |||
Resolved | PRODUCTION ERROR | Andrew | T168470 Setup wikitech, horizon, and striker on new labweb hardware | ||
Resolved | PRODUCTION ERROR | Krinkle | T143464 "Memcached::touch(): touch is only supported with binary protocol" from wikitech as it's not running HHVM | ||
Resolved | Andrew | T98813 Move wikitech (silver) to HHVM |
Event Timeline
See also T87036 - although silver runs trusty and has PHP 5.5 rather than 5.3, we should still migrate it to HHVM.
This ticket makes no sense, IMO. We have no reason to migrate wikitech to HHVM at the moment.
I'm resolving this ticket as invalid until someone explains me why should we embark in such a migration that - at the moment - does not offer any advantage to us.
It's an absolute blocker for T91590: [Spike] Try out hack (<?hh) for mediawiki-config, T94149: Get rid of Zend 5.5 tests for wmf branches, and perhaps T75901: Drop PHP 5.3 support. If you want to close this, you'll have to reject at least those first two.
This is not a blocker to drop php 5.3, it is not a blocker to drop zend tests (or are we testing OSM? we might keep tests just for it)
Converting silver could be a 5-minutes job or an engineering nightmare requiring weeks to fix OSM to work correctly with HHVM. I strongly suspect the latter.
If we really want to have an HHVM-only release train, we could re-detach wikitech until we drop OSM.
@andrewbogott: thoughts?
I have no idea whether hhvm will interact badly with OSM. It's pretty different from other extensions so it wouldn't shock me. A day long wikitech outage isn't a huge deal since it doesn't prevent actual labs instances from working -- we'd need to schedule it ahead of time though.
If it were detached from the train would I still be able to manually sync-common or would I be back to upgrading it fully by hand?
As there is some disagreement here, trying to be neutral here and just reflecting the current state of this task.
I removed OSM deprecation as a blocker. I think we can figure out how to work through any issues that OSM has with HHVM. Its mostly LDAP and cURL.
See T168470. Cloud Services has 2 new physical servers that were purchased to replace the aging silver (wikitech) and californium (horizon, striker) hosts. The rough plan is for these new servers to be built out as Debian boxes and setup with all 3 services on each physical node. Then we can balance traffic between them and make maintenance things easier. We should have the luxury of setting these boxes up in parallel with the existing hosts, so in my mind that would be a good time to attempt to run wikitech's weird mix of extensions on HHVM and make sure that the obvious problems get fixed. This work may become a FY17/18 Q2 goal for the team.
The broad plan for PHP7 is to move to HHVM on Debian Stretch first and then move from HHVM to PHP7. Getting wikitech running on Stretch+HHVM will put us in a good place to continue on to PHP7 with the rest of the MW servers.
Wikitech on Silver has been replaced by Wikitech on labweb1001/1002. The new wikitech uses HHVM.