[06:14:50] arturo: right [11:19:53] bd808: do we have the WMCS IP ranges so I can whitelist them from autoblock? [11:25:55] Hauskatze: he is on vacations [11:26:23] arturo: ah, ok :) [11:26:40] (and he keeps on irc? :) ) [11:26:47] Hauskatze: this might help you https://wikitech.wikimedia.org/wiki/IP_and_AS_allocations [11:27:00] what is autoblock, BTW? [11:27:02] arturo: gracias, checking [11:27:15] arturo: a mediawiki blocking tool [11:27:36] https://www.mediawiki.org/wiki/Autoblock [11:28:38] ok, so we are only interested in whitelisting toolforge? [11:29:49] Hauskatze: this might help even more: https://tools.wmflabs.org/openstack-browser/project/tools list of FQDNs, public IP addresses and private addresses from toolforge. A given tool/bot running on toolforge will likely look as coming from there [11:30:16] arturo: point is to avoid bots from being IP blocked in case of "rogue" blocks [11:30:46] ack [11:30:55] ok so that's 80.208.155.0/24 [11:32:33] well, according to the first wiki link, it's `208.80.155.128/25 - labs virtualization - floating IPs` [11:33:48] but I would say some contacts may be made using private IPs? not sure [11:35:03] .0/24 is still wmf so no problems I'd say [11:35:24] 208.80.155.0/24 - eqiad [14:36:26] Technical Advice IRC meeting starting in 30 minutes in channel #wikimedia-tech, hosts: @Tonina_WMDE & @Christoph_Jauera_(WMDE) - all questions welcome, more infos: https://www.mediawiki.org/wiki/Technical_Advice_IRC_Meeting [15:19:52] !help I created a new project half an hour ago (https://toolsadmin.wikimedia.org/tools/id/kokolores) but I still get "no such tool" when trying to become the tool, even after logging out and in multiple times [15:19:52] CirdanWP: If you don't get a response in 15-30 minutes, please create a phabricator task -- https://phabricator.wikimedia.org/maniphest/task/edit/form/1/?projects=wmcs-team [15:20:11] Let me take a look at that. [15:20:45] On Phabricator I saw some older tasks with similar problems, from which it seems that there is a service which sometimes crashes (maintain-kubeusers on tools-k8s-master-01) [15:20:56] bstorm_: thanks! [15:21:18] Well, depending on which tasks, it may have been around our network maintenance where we crashed it pretty firmly :) [15:21:36] It is sometimes a tad slow either way [15:22:12] yeah, I'm fine with waiting a bit longer, I just thought I'd come here instead of filing a potentially unnecessary Phabricator task [15:22:35] I'm looking at the service and such. It definitely hasn't made your directory for that yet [15:23:46] CirdanWP: hi [15:25:16] Cyberpower678: Hi :) [15:25:30] CirdanWP: you up for Skype? [15:42:53] CirdanWP: the service appears to be running normally, but it never found the new name. The tools is in LDAP now, which means the service should find it, but it hasn't. I restarted the service (and even tried creating the folder manually--which will work, but it will be missing some things). I'll keep poking a bit. [15:43:20] thanks for working on this bstorm_ [15:43:27] bstorm_: Thanks for looking into this! [15:43:53] If it isn't created shortly (I removed my manually set up one), we should probably make a task. I'll also see if there is something I'm missing somehow around LDAP. [15:49:15] bstorm_: Thanks. I'll have a look later today and if it still hasn't been created, I will open a task [15:49:40] In the meanwhile I'll keep looking around in case there is something else I can do or something else that isn't working that we haven't thought of. [16:22:04] bstorm_: I just saw that while poking around to figure out why my denkmalbot project was showing but not kokolores, I removed the toolforge prefix: https://toolsadmin.wikimedia.org/tools/id/kokolores/info/id/335/history?version_id2=448&version_id1=447 [16:22:24] bstorm_: I reverted the change, but perhaps that messed with the creation process? [16:23:08] It doesn't seem like it should be it. This polls LDAP every minute and should recognize tools that it doesn't know. [16:23:18] Buuuuut [16:23:21] I'll check that just in case [16:33:50] Heh. Interesting thing is that running the exact code piece by piece, this particular tool id does not come out of LDAP. I suspect there is something peculiar about it. I'll confirm [16:33:58] So you might have the reason [16:34:06] the toolforge prefix [16:34:16] or revert or something [16:34:19] poking around [16:39:11] I have a weird idea about this. Recently, I created a tool to test and it took FOREVER to create it. I'm wondering if there isn't something going on with the timelimit used in the code for querying LDAP [16:39:44] I am able to make the kokolores tool show up in LDAP easily, and it all looks correct, regardless of anything you may have done. That means it should work. [16:43:04] bstorm_: it might also be worth keeping an eye out for ldap page boundaries. That catches us up from time to time. [16:43:46] Thanks. [16:44:00] I had to mess with that to get my ldapsearch queries right to confirm things :) [16:52:16] If this responds to the ldap python module anything like it responds to ldapsearch, this needs changes to work with paging. I'll confirm that with my testing script [16:59:01] bstorm_: I'm off now for a bit, I'll check in again in about an hour and a half. Thanks a lot for investigating this issue! [16:59:44] Fixed it in my test script! [16:59:58] It needs the python module to actually use paging and resend the cookie and all that [17:00:04] I'll make a patch. [17:00:24] np CirdanWP. I'm glad you brought it up! This was broken. [17:15:41] may I see the patch? I'm curious [17:17:02] Yup...submitting it now [17:18:19] https://gerrit.wikimedia.org/r/434744 [17:19:55] It's just adding pagination to the ldap search (which is apparently needed in my testing), and a few python nitpicks [17:24:36] .... how did you know it was there? :-P that's totally in the shadow for me [17:25:41] lots of poking around? [17:26:22] As is, it's totally luck if a new tool gets created by virtue of random ordering. This causes the script to get all the tools every time [17:28:33] also, andrewbogott, thank you, that's exactly what was up. [17:30:57] arturo: if you have a moment before you fall asleep to look at my patch, I'm curious if I appear to have done anything silly. Most of the reviewers are away :) [17:31:18] * arturo reading [17:36:47] bstorm_: I can't comment on algorithm to paginate the LDAP query result, I've never done that before. However it makes sense. The patch itself could be improved, also python code: doing unrelated changes in the same patch, string formatting, etc [17:37:31] Yeah...but the original code has so many problems! [17:37:35] ;-) [17:38:03] I know :-) [17:38:19] other than that, it looks great! [17:38:54] as I said... it seems magic to me you are able to patch this :-P [17:39:19] * arturo knows so little about toolforge internals/k8s...and many other stuff.... [17:39:55] I'll split out my nitpicks (a couple of which have performance implications), and update it to be strictly the fix here. Then I'll merge it, and push a new one with just my nitpicks (so other people can add theirs 😁) [17:41:17] ack [17:46:15] The great thing is that the old code was not entirely working. The beautiful thing about working on broken code is that you cannot do *that* much harm ;-) [17:52:16] CirdanWP: it's fixed. This also fixed two other attempted account creates. [17:52:39] Now to put up my nitpick patch :-) [17:59:48] bstorm_: If necessary we can just make a quick hack to increase the page limit. I'm not really following what all you're up to though, maybe there's a better fix [18:01:33] I page through it in the code. That ensures I get all entries and obey api rules [18:01:51] ok :) [18:02:03] worked great [18:02:46] of course, it may be needed in at least one more script, then. I will check for that. [19:10:07] bstorm_: thanks a lot for bug hunting and fixing, the kokolores project is ready to go! [19:10:36] :) [20:28:32] !log discourse ran docker image prune [20:28:33] Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Discourse/SAL [20:28:45] !log discourse ran ./launcher rebuild app [20:28:46] Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Discourse/SAL [21:09:49] hieradata/labs.yaml has `elasticsearch::version: 2` in it, is that actually being used? I was going to clean out old v2 puppet code but if it's still used i'll leave it alone