Hi Andrew,
Andrew Bogott schreef op 16-5-2015 om 6:31:
I did shut off one instance: wikidata-wdq-mm. I don't have a personal grudge, but it was gobbling CPU cycles and the system really needs a rest. If loss of that instance is a disaster for anyone, contact me and I'll see if I can revive it and shut off ten or so other instances to make room.
With that you basically break the edit flow of most users on Wikidata, see https://www.wikidata.org/wiki/Wikidata:Project_chat#wdq.wmflabs.org.2Fapi . This is one of those tools that have silently become production.
Maarten
On Sat, May 16, 2015 at 4:02 AM, Maarten Dammers maarten@mdammers.nl wrote:
With that you basically break the edit flow of most users on Wikidata, see https://www.wikidata.org/wiki/Wikidata:Project_chat#wdq.wmflabs.org.2Fapi . This is one of those tools that have silently become production.
Nope - wdq.wmflabs.org runs from wdq-mm-01.eqiad.wmflabs and wdq-mm-02.eqiad.wmflabs. wikidata-wdq-mm is just the old instance, currently unused (afaict)
On Sat, May 16, 2015 at 4:02 AM, Maarten Dammers maarten@mdammers.nl wrote:
With that you basically break the edit flow of most users on Wikidata, see https://www.wikidata.org/wiki/Wikidata:Project_chat#wdq.wmflabs.org.2Fapi . This is one of those tools that have silently become production.
It may not be correct to say "you" here :). If something is really important, it should be run in a way that can handle failure of the underlying hardware. Labs infrastructure wasn't designed for high-uptime of the underlying instances (purposely).
If it's production-ish, it should likely either be moved to production or you should put a bit of effort into making it work across multiple instances. The ideal goal is for services to be stateless, with their state living in databases that are also split across instances. It's best to have the service config managed (ideally puppetized since it's what wikimedia uses) so that a loss of an instance is only a brief inconvenience.
- Ryan
Ryan Lane rlane32@gmail.com wrote:
[WDQ]
If it's production-ish, it should likely either be moved to production or you should put a bit of effort into making it work across multiple instances. The ideal goal is for services to be stateless, with their state living in databases that are also split across instances. It's best to have the service config managed (ideally puppetized since it's what wikimedia uses) so that a loss of an instance is only a brief inconvenience.
There are efforts to deploy a similar service with https://www.mediawiki.org/wiki/Wikibase/Indexing (Phabrica- tor project at https://phabricator.wikimedia.org/tag/wikidata-query-service/).
Tim