Uhm, tech still lists OTRS on iodine, which seems to be decommissioned. I had a look at http://ganglia.wikimedia.org/latest/?r=year&cs=&ce=&m=cpu_repor… and...well fairly busy but still "packable", though I'll let more experienced people think about it ;) Vito Il 04/04/2016 20:11, Alex Monk ha scritto:

I would like to see Phabricator move to a virtual owed cluster, or even a bare metal cluster. However, I would not include Phabricator in a list of "not-so-resource-consuming services." Phabricator definitely uses significant amount of resources. It's running on a 12 core server with 64gb ram and 500gb raid storage, currently. That hardware is a little bit more than required but not by very large margin. I've recently requested a second machine to be used for a backup Phabricator instance, so that we can avoid long downtimes during maintenance or hardware failure. That request is still being discussed at https://phabricator.wikimedia.org/T131775 Phabricator is gaining improved high-availability support thanks to recent work upstream, so it might be possible to have dual-master phabricator nodes in the near future. See https://secure.phabricator.com/T10751 for upstream progress. On Monday, April 4, 2016, Vituzzu <vituzzu.wiki(a)gmail.com> wrote:

Indeed, Jaime, I think that will be possible very soon. The upstream task I linked to previously (https://secure.phabricator.com/T10751) is about two things: 1. properly handling clustered databases, especially dealing with read-only slaves 2. Implementing replicated git hosting, where repositories can exist on multiple back-end nodes with reads going to any up-to-date copy of the repo and writes replicated to every other copy. It seems like progress is being made on both fronts and I am happy to see it. On Fri, Apr 15, 2016 at 12:35 AM, Jaime Crespo <jcrespo(a)wikimedia.org> wrote: