[01:37:41] hello! I've got my first bot task written in Node. When I try to run it with jsub it always fails with "Fatal process OOM in heap setup". Running the script directly works, however. Search engines suggest this has something to do with ulimit? I tried giving the job more memory [05:40:33] Hi. any one here? [05:40:46] I have a bot in the toolserver. It uses nodejs to run the task. [05:40:47] 05:05:04 1 [05:40:50] still hanged. [05:40:53] know what's going wrong. May someone help me? [06:24:12] https://phabricator.wikimedia.org/T229084 [06:38:51] I have a bot in the toolserver. It uses nodejs to run the task. [15:54:59] !log tools created tools-worker-1031 - T228573 [15:55:03] Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools/SAL [15:55:03] T228573: toolforge k8s nodes oom? - https://phabricator.wikimedia.org/T228573 [15:57:01] !log tools created tools-worker-1032 and 1033 - T228573 [15:57:04] Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools/SAL [16:32:21] !log tools created tools-worker-1034 - T228573 [16:32:25] Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools/SAL [16:32:25] T228573: toolforge k8s nodes oom? - https://phabricator.wikimedia.org/T228573 [16:44:32] interested another user was having trouble with a Node.js bot! but they didn't have the same problem as me, it seems [16:44:49] I'll ask agian since many of you are awake now: I'm getting "Fatal process OOM in heap setup" when I try to run my Node script via jsub [16:45:00] *interesting [16:45:40] but I can run the script directly (which I know you're not supposed to do) [16:46:38] all the docs on Wikitech are about Node.js webservices, nothing about running them via cron. Surely there's a way to do this? [17:09:24] !log tools draining tools-worker-1020.tools.eqiad.wmflabs to rebalance load [17:09:27] Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools/SAL [17:10:41] musikanimal: what vmem value are you giving the jsub job? [17:11:14] memory? I tried as much as 500m. I don't think it's a memory issue, as it's a fairly lightweight job [17:12:45] it runs a quick query with 15-20 results, then updates a page, that's it [17:13:32] https://github.com/DannyS712/bot (polluted-category.js, specifically) [17:13:42] node does a lot of overcommitting to make sure it will have enough memory in the future, which is why I thought of that [17:14:06] I'd give it a gb and see if that helps. There have been memory issues on the cron host in the past as well, so I might check that [17:14:13] I'll check that in a few minutes [17:14:57] !log tools drained tools-worker-1013.tools.eqiad.wmflabs to rebalance load [17:14:59] Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools/SAL [17:15:12] whaaa, okay it's running with 1g of memory! [17:15:49] interpreted languages are silly sometimes [17:16:03] I doubt it actually uses a tenth of that [17:16:11] it can't [17:16:15] heh [17:16:36] but alright, I guess this will work... haha. Thank you! [17:17:13] np :) [17:17:25] an unrelated problem, this user that I'm helping created a new tool account because they didn't want the old name [17:17:27] https://toolsadmin.wikimedia.org/tools/id/dannys712-bot [17:17:44] but `become dannys712-bot` doesn't work for either of us, says no such tool exists [17:22:07] Hrm. I wonder if something didn't survive NFS maintenance. Checking [17:25:07] it was created about 13 hours ago [17:26:35] become: no such tool 'dannys712-bot' hrm [17:30:26] It could have been done during nfs maintenance. I'll check some other things. If so, I can manually create the NFS dir, and the other bits should fall in place. [17:31:25] yeah, it's in ldap, so it looks that way [17:36:33] musikanimal: works now. It hasn't created kubernetes credentials yet, but it has DB creds already [17:36:44] awesome! thank you :) [17:39:03] There, it has those creds now as well [17:39:38] !log tools restarted maintain-kubeusers because it was suspiciously tardy and quiet [17:39:41] Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools/SAL [17:56:41] bstorm_: ugh. maintain-kubeusers blew up last week too. [17:57:17] I've found it does that periodically. However, with NFS maintenance, I should have figured it was guaranteed to. [17:57:42] *nod* that could freak it out for sure [17:58:23] I'd take a look at making it more robust again, but I think you are more excited about replacing it with another thing entirely "soon" [20:54:01] Hi everyone! Would any of the bot/ tool developers who are familiar with the Action API be willing to review these short code snippets in https://phabricator.wikimedia.org/T228549? The goal is to show them in a format like this https://test.wikipedia.org/wiki/API:Sandbox on the API module pages on MediaWiki.org. This work is part of my GSOC project. Your help would be much appreciated. If you are interested [20:54:01] , feel free to claim the task on Phabricator. [21:08:03] Jayprakash12345: for the PHP ones, you'll want to set a useragent as one of the curl options to properly identify your tool [21:08:34] I'll admit I've not looked at the others to see if they have the same point :) [21:10:10] great feedback Oresrian. Having a custom User-Agent header that gives some way of contacting you (wiki talk page, email, etc) is a best practice for bot activity against Wikimedia APIs [21:13:36] Oresrian: Thank you for your great suggestion :) [21:18:31] Jayprakash12345: if you don't get as much feedback as you would like, you might try asking on the mediawiki-l and/or cloud mailing lists too. [21:19:03] bd808: Ok Sir :) [21:19:41] Jayprakash12345: how is this project going for you? Are you enjoying the process? [21:23:24] bd808: Very good sir and I am enjoying very much, I love Wikimedia Technical community very much that is why I am still a part of the Wikimedia community since 2015. [21:24:02] :)