[12:46:52] !log toolsbeta Joining grid node toolsbeta-sgewebgen-10-1.toolsbeta.eqiad1.wikimedia.cloud to the toolsbeta cluster - cookbook ran by arturo@endurance [12:46:54] Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Toolsbeta/SAL [12:55:03] !log paws deploying dcac479b8c68631977441be5aaff8f6335659e1a T294431 [12:55:09] Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Paws/SAL [12:55:09] T294431: requirements.txt for renderer - https://phabricator.wikimedia.org/T294431 [16:55:02] Hi everybody [16:55:30] Got a WMF banned user abusing PAWS 2.1 and pywikibot to make vandal edits on es.wikipedia [16:55:47] (For reference: https://meta.wikimedia.org/wiki/Special:Diff/22457701 ) [16:56:15] Is there anything that can be done to stop such abuses, perhaps by preventing it from being used by users who are not auto-confirmed? [17:27:56] LuchoCR: its not an easy thing to fix, but we are aware that there are some ways that PAWS can be abused. It does turn out however that any abuse that can be done via PAWS can also be done other ways, meaning that there is no special path for abuse that PAWS creates. [17:39:35] LuchoCR: you can CU 'em to see where they signed up from [17:43:08] Spain; and proxis from Afghanistan, Ghana and Ivory Coast [17:43:17] proxys* [19:30:27] hi, what is the timeout of normal queries in toolforge (and quarry?) [19:30:33] db queries to the replcias [19:57:25] Is it possible to attach a volume to an instance at the time of creation, or to otherwise use the customization script to take care of configuring a Cinder volume? [19:58:01] or can that only be done manually from Horizon after an instance is done launching? [20:01:36] ragesoss: I think we have only figured out how to do it manually so far. andrewbogott would be the one who would know differently if a fully automated way is possible that I've missed hearing about. [20:02:19] cool, thanks bd808. [20:02:26] ragesoss: currently only manual. OpenStack/Horizon in theory supports automatic creation/attachment of cinder volumes but last time I tried it it didn't work at all so I disabled the UI [20:02:41] If you have some repeating workflow that needs this we could investigate it further. [20:03:12] It's also possible to create a flavor with an attached 'ephemeral' drive on creation. That drive, though, will be destroyed when the VM is destroyed and can't be detached/moved to other VMs. Hence 'ephemeral' [20:03:35] we've found that it's only really appopriate for niche uses but if you have one I can set you up with a custom flavor that includes ephemeral storage. [20:03:57] no, I'm just working on figuring out the process for setting up this WikiWho server, and I've found a lot of it really suitable to putting into the customization script. [20:04:27] but it's not actually important to automate it... just wanted to keep as much in the script and out of manual instructions as I can manage. [20:05:54] ok. Especially with a huge volume that you might want to reattach to a new VM, it's probably good to keep volume creation as a separate/manual step [20:06:02] like, automatically installing and enabling an nginx conf via customization script... awesome! wish I'd played around with this approach earlier. [20:06:12] yeah, this is a huge volume... 5TB! [20:07:37] I was hoping, rather, for a way to attach an already-existing volume via script. [20:08:09] Amir1: I think I heard a Quarry timeout of 30 minutes at some point [20:08:26] Thanks [20:08:34] I don’t know if that’s specific to Quarry or the cloud DBs in general [21:05:00] Amir1: when things are working 100% normal, the *.web wiki replicas have a 5 minute timeout before the query killer can stop a query and the *.analytics wiki replicas have a 180 minute timeout. [21:05:21] thanks! [21:05:54] Sometimes the 5 minute timeout applies on the *.analytics nodes though if we are doing some kind of maintenance and connecting both *.web and *.analytics to the same single instance. [21:06:27] Quarry always sends its queries to the *.analytics instances. [21:07:30] Quarry used to have its own long running query killer too, but I don't know if it still does...