[16:20:42] MediaWiki 1.16.5 http://linux-nfs.org/wiki/index.php/Special:Version [16:59:15] <_joe_> 5.4.16 (apache2handler) [17:09:46] Oh dear. On both counts. [19:19:51] robh: hi. I think it's hard sell but I was hoping to interest you in T238652 :) [19:19:54] T238652: Hardware request for Postgres database for censorship monitoring scripts - https://phabricator.wikimedia.org/T238652 [19:20:05] s/it's/it's a [19:20:19] sukhe: So this is a request for a previously non-budgeted new server? [19:20:39] I am not sure if "non-budgeted" or not but this is a request for a new server, yes. [19:20:53] hrmm [19:20:55] ' CPU/RAM/Storage: Aside from the need for ~1TB of disk space, the rest seem, assuming the usual reasonable ratios of CPU/RAM we have across the fleet seem fine. [19:21:01] what fleet? [19:21:22] I mean, it isn't clear to me if this is going to need the DB class machine or just a random box =] [19:21:27] sorry, akosiaris helped me draft that but I am not sure if he is around [19:21:31] hrmm [19:21:35] yeah, just thining aloud [19:21:39] seems this is ONLY hardware due to storage space [19:21:45] so a single CPU misc server may work [19:21:45] =] [19:22:13] that is my understanding as well. otherwise a VM would have been sufficient [19:22:17] sukhe: i'll comment on task yep [19:22:20] thank you! [19:22:26] checking to see if we have a spare sysetm in eqiad that would work before i do [19:23:26] bleh. all our spares in eqiad have ssds [19:23:33] and they aren't larger than 1TB, they're like dual 480s [19:24:43] current size of the db is: 720Gtotal [19:25:26] yeah, so our single cpu class machine [19:25:33] but swap dual 480GB ssd for dual 2TB SATA [19:25:36] and we should be good. [19:25:52] its that or scale up lots of SSDs and that is more expensive [19:26:09] and ssd speed for this doesn't exactly seem important [19:26:13] where not running out of space is! [19:26:38] not at all, speed is not important. we query the database once everyday and that's about it, while running some queries in between [19:26:56] currenlty it's running on my old laptop hooked to an external HDD :) [19:27:06] cool, i have a quote back similar to this to proof, and if it checks out ill use it as a baseline for your request and send the quote requests in later today [19:27:34] thanks. sorry for not being helpful with the other stuff but I am happy to ask bblack if required [19:28:13] I think the request even said spares were ok, but maybe we don't have spares with space [19:29:54] sukhe: no worries [19:30:00] sukhe: can you let me know if you can see https://phabricator.wikimedia.org/T239156 ? [19:30:13] if not ill add you to the proper groups [19:30:23] bblack: indeed no spares in eqiad with that space [19:30:27] as spares now have dual 480GB ssd now [19:30:33] but otherwise a single cpu system will work [19:30:34] robh: I can access it [19:30:38] do we ever reconfig spares? [19:30:47] we lose warranty on whatever we swap [19:30:51] and its hard to track [19:30:54] so we try not to unless we have to [19:30:54] (like, take a spare and buy a pair of 2TB cheap disks to stuff in it) [19:31:02] but yeah, we totally have in the past [19:31:13] the issue is the spindle count cannot change [19:31:17] or else we have to order the disks from dell ;D [19:31:17] ok [19:31:20] (so we have caddy) [19:32:24] bblack: though good piont and ive noted it as option now on the procurement task [19:32:35] cmjohnson1: do you guys have 2tb sata disks in SFF spare by chance? [19:32:41] if they do, we could get this done this week ;D [19:33:10] i mean, its likely better to use a old spare thats already a year old [19:33:21] and swap hardware than use money on new host for a low performance requirement [19:33:54] also this doesn't need to be reliable as we are making read-only queries to the db. (I don't want it to fail as well but if it does, it's not a big deal!) [19:34:12] oh, then i really rather use existing spare [19:34:15] s/reliable/critical [19:34:20] sukhe: that is super helpful thank you [19:35:34] robh: yeah, it's not running critical code as in case it fails (I hope it doesn't), we can sync it again and also make queries against the API. it's just having the db itself is very useful for quick access [19:35:52] the API is limited and the db is not [20:12:44] a virtual machine is not an option for this case? [20:13:54] mutante: I was told that is not because of the disk space requirements [20:14:36] ah, gotcha