| *** mralex has quit IRC | 00:00 | |
| *** JoiIto has joined #cc | 00:23 | |
| *** sudiptamondal has joined #cc | 00:35 | |
| *** JoiIto has quit IRC | 00:36 | |
| *** sudiptamondal1 has joined #cc | 00:39 | |
| *** luisv has quit IRC | 00:40 | |
| *** sudiptamondal has quit IRC | 00:40 | |
| *** akila87 has joined #cc | 00:43 | |
| *** JoiIto has joined #cc | 00:44 | |
| *** JoiIto has quit IRC | 00:48 | |
| *** MarkDude has quit IRC | 01:04 | |
| *** igorlukanin has joined #cc | 01:20 | |
| *** igorlukanin1 has joined #cc | 01:21 | |
| *** igorlukanin1 has quit IRC | 01:36 | |
| *** nkinkade has quit IRC | 01:52 | |
| *** JoiIto has joined #cc | 01:59 | |
| *** oshani has quit IRC | 02:17 | |
| *** JoiIto has quit IRC | 02:29 | |
| *** sudiptamondal1 has quit IRC | 02:35 | |
| *** Kaetemi has quit IRC | 03:23 | |
| *** Kaetemi has joined #cc | 03:23 | |
| *** JoiIto has joined #cc | 04:13 | |
| *** JoiIto has left #cc | 04:17 | |
| *** akozak has quit IRC | 05:36 | |
| *** Kaetemi has quit IRC | 06:04 | |
| *** sudiptamondal has joined #cc | 06:33 | |
| *** wormsxulla has quit IRC | 07:27 | |
| *** michi_ has joined #cc | 07:28 | |
| *** michi_ has quit IRC | 07:34 | |
| *** sama_ has joined #cc | 07:41 | |
| *** wormsxulla has joined #cc | 07:44 | |
| *** sudiptamondal has quit IRC | 08:35 | |
| *** sudiptamondal has joined #cc | 08:56 | |
| *** erlehmann has joined #cc | 08:59 | |
| *** anttil has quit IRC | 09:01 | |
| *** akila87 has left #cc | 09:09 | |
| *** anttil has joined #cc | 09:13 | |
| *** wormsxulla has quit IRC | 09:51 | |
| *** wormsxulla has joined #cc | 10:07 | |
| *** anttil has quit IRC | 10:09 | |
| *** akila87 has joined #cc | 10:10 | |
| *** bassel has joined #cc | 10:23 | |
| *** bassel has quit IRC | 10:42 | |
| *** sudiptamondal has quit IRC | 11:21 | |
| *** anttil has joined #cc | 11:31 | |
| *** user01 has joined #cc | 12:02 | |
| *** user01 has quit IRC | 12:04 | |
| *** user01 has joined #cc | 12:05 | |
| *** user01 has quit IRC | 12:11 | |
| *** Roux_taff has joined #cc | 12:57 | |
| *** Roux_taff has left #cc | 12:57 | |
| *** tvol has joined #cc | 13:18 | |
| *** everton137 has joined #cc | 13:25 | |
| *** oshani has joined #cc | 13:25 | |
| *** sama_ has quit IRC | 13:35 | |
| *** sama_ has joined #cc | 13:35 | |
| *** sama_ has quit IRC | 13:37 | |
| *** sama_ has joined #cc | 13:38 | |
| *** everton137 has quit IRC | 13:41 | |
| *** sama has joined #cc | 13:58 | |
| *** sama_ has quit IRC | 14:01 | |
| greg-g | JED3: question about the humanizer, I thought I had the morepermissions stuff correct on my test page, but the humanizer is giving "Permissions beyond the scope of this public license are available at ." | 14:03 |
|---|---|---|
| greg-g | url: http://grossmeier.net/files/test.html | 14:03 |
| *** akila87 has quit IRC | 14:08 | |
| *** akila87 has joined #cc | 14:10 | |
| *** Odemia has quit IRC | 14:12 | |
| *** Odemia has joined #cc | 14:13 | |
| *** Oxygeek has joined #cc | 14:17 | |
| *** nkinkade has joined #cc | 14:19 | |
| *** Oxygeek has left #cc | 14:20 | |
| *** sama_ has joined #cc | 14:40 | |
| *** sama has quit IRC | 14:43 | |
| *** akila87 has left #cc | 14:52 | |
| *** mattl_ is now known as mattl | 14:56 | |
| *** mattl has joined #cc | 14:56 | |
| *** Kaetemi has joined #cc | 14:56 | |
| *** sudiptamondal has joined #cc | 15:06 | |
| *** oshani has quit IRC | 15:11 | |
| *** oshani has joined #cc | 15:12 | |
| *** jgay has joined #cc | 15:15 | |
| *** sama_ has quit IRC | 15:16 | |
| *** sama_ has joined #cc | 15:16 | |
| *** sama_ has quit IRC | 15:28 | |
| *** oshani has quit IRC | 15:33 | |
| *** akila87 has joined #cc | 15:34 | |
| *** sudiptamondal1 has joined #cc | 15:34 | |
| *** oshani has joined #cc | 15:34 | |
| *** sudiptamondal has quit IRC | 15:36 | |
| *** sudiptamondal1 has quit IRC | 15:39 | |
| *** sudiptamondal has joined #cc | 15:41 | |
| *** oshani has quit IRC | 15:46 | |
| JED3 | nkinkade: ping | 15:55 |
| nkinkade | JED3: Here. | 15:56 |
| JED3 | could i solicit your help for just a sec? | 15:56 |
| *** erlehmann has quit IRC | 15:57 | |
| JED3 | the scraper went down from what looks like when the logs were rotated, creating new empty log files with invalid permissions | 15:57 |
| JED3 | nkinkade: do you know much about the SuexecUserGroup configuration for mod_fcgid? | 16:00 |
| nkinkade | JED3: No, I know little to nothing about the SuExec stuff. | 16:01 |
| nkinkade | So does the scraper need to be restarted after the logs are rotated? | 16:01 |
| *** JoiIto has joined #cc | 16:01 | |
| JED3 | nkinkade: when the logs are rotated they are set with permissions that the scraper cannot write to them | 16:02 |
| nkinkade | JED3: Logrotate can handle that and set the new log with certain perms. | 16:03 |
| JED3 | the current SuexecUserGroup is set to "john john" | 16:03 |
| nkinkade | Check the "create" directive of logrotate. | 16:03 |
| nkinkade | create mode owner group | 16:03 |
| nkinkade | JED3: What do the perms need to be, and owner/group? | 16:04 |
| JED3 | nkinkade: does it make more sense to fix this in the logrotate or changed the suexecusergroup to something like "www-data www-data" | 16:04 |
| nkinkade | JED3: If you are happy and comfortable with the current user/group of the log files, from a security standpoint, then I'd say we just use logrotate to fix it. | 16:05 |
| JED3 | nkinkade: i'd prefer that the logs not be owned by 'john' | 16:06 |
| nkinkade | JED3: Should root own them? | 16:06 |
| JED3 | same for the scraper's fcgi process | 16:06 |
| nkinkade | JED3: What should the fcgi process run as? | 16:06 |
| JED3 | i'm going to change the suexecusergroup then | 16:06 |
| JED3 | www-data users? | 16:06 |
| nkinkade | Seems to me it's better john, than root. | 16:06 |
| nkinkade | Oh, www-data could be good, I think. | 16:07 |
| nkinkade | www-data doesn't own much. | 16:07 |
| JED3 | oh but www-data wont work will it? does the user need to be a sudoer? | 16:07 |
| nkinkade | JED3: I don't think so. | 16:07 |
| nkinkade | SuExec will just run the process as the owner of the script, right? | 16:08 |
| JED3 | nkinkade: not sure, i dont know much about suexec | 16:08 |
| nkinkade | JED3: I'm pretty sure that's how it works. If www-data owns the script, then the resulting process will be run as user www-data. Though I could be wrong. Try it. | 16:09 |
| *** nyergler has joined #cc | 16:09 | |
| JED3 | okay | 16:09 |
| *** oshani has joined #cc | 16:22 | |
| JED3 | nyergler: ping | 16:22 |
| *** akila87 has left #cc | 16:22 | |
| *** akozak has joined #cc | 16:26 | |
| *** luisv has joined #cc | 16:28 | |
| nyergler | JED3, pong | 16:29 |
| JED3 | nyergler: hi, i'm trying to set where i want the python-egg-cache to be using buildout, but the config doesn't seem to be taking | 16:30 |
| JED3 | nyergler: is the configuration option "environment-variables" or "environment" ? | 16:31 |
| nyergler | environment, i think | 16:31 |
| nyergler | JED3, how are you trying to set it? | 16:32 |
| JED3 | neither have worked for me... | 16:32 |
| JED3 | environment = | 16:32 |
| JED3 | PYTHON_EGG_CACHE ${buildout:directory}/var/.python-eggs | 16:32 |
| JED3 | nyergler: ^^ | 16:33 |
| nyergler | JED3, in the zc.recipe.egg section? | 16:34 |
| JED3 | yes, tried it there and in the main buildout part | 16:34 |
| nyergler | JED3, looks like you have the format wrong | 16:35 |
| nyergler | see http://pypi.python.org/pypi/zc.recipe.egg#controlling-environment-variables | 16:35 |
| nyergler | JED3, looks like you need to use the "custom" entry point for the recipe, too | 16:36 |
| nyergler | JED3, possibly unrelated, did you see my email re: the scraper? | 16:36 |
| JED3 | nyergler: directly related to my questioning :) | 16:37 |
| nyergler | great | 16:37 |
| JED3 | the logs were rotated with different permissions, decided to change the suexecusergroup of the scraper fcgid to www-data, can't run the scraper as it tries to create a python-egg cache in the /var/www/ directory which is the home of the newly appointed user, trying to change that now. | 16:39 |
| JED3 | ^^ which is why i'm asking buildout questions | 16:39 |
| nyergler | JED3, ah | 16:40 |
| nyergler | JED3, I really hate egg caches | 16:41 |
| nyergler | JED3, why not just set unzip=true in buildout, blow away your eggs, re-run buildout so that eggs are unzipped, and ignore the cache? | 16:41 |
| nyergler | (of course I'd suggest backing up your current deployment just in case something goes horribly, unexpectedly awry). | 16:42 |
| nyergler | JED3, but regardless, it looks like you have to specify environment as a buildout section, not as inline variable assignments | 16:43 |
| JED3 | nyergler: trying unzip now | 16:43 |
| JED3 | nyergler: yeah but when i used the egg:custom recipe didn't seem to like how i was declaring the egg dependencies | 16:43 |
| *** mralex has joined #cc | 16:51 | |
| *** bassel has joined #cc | 17:00 | |
| nyergler | JED3, did unzip work? | 17:07 |
| JED3 | nyergler: yes that did work, but now i'm chasing down why fcgid is sending out 503's | 17:07 |
| JED3 | i can run the server as www-data (same user thats declared in the suexecusergroup config) in the foreground just fine | 17:09 |
| *** sudiptamondal has quit IRC | 17:18 | |
| *** oshani has quit IRC | 17:29 | |
| *** sudiptamondal has joined #cc | 17:37 | |
| *** sudiptamondal has quit IRC | 17:49 | |
| *** jgay has quit IRC | 17:54 | |
| *** jgay has joined #cc | 17:54 | |
| *** jgay has joined #cc | 17:55 | |
| *** jgay has joined #cc | 17:57 | |
| *** jgay has quit IRC | 17:58 | |
| nyergler | JED3, if you haven't already moved on, might try turning up the log level to get more details about the problem | 17:58 |
| *** sama_ has joined #cc | 17:58 | |
| *** jgay has joined #cc | 17:58 | |
| *** jgay has joined #cc | 17:58 | |
| nkinkade | JED3: mralex: This page on the CC Wiki is hanging, and eventually returns a 503 from Varnish: | 17:59 |
| nkinkade | http://wiki.creativecommons.org/Special:NewFiles | 17:59 |
| nkinkade | The very odd thing is that it's not even getting far enough for Apache to log the request. | 18:00 |
| JED3 | nyergler: any idea what this is "mod_fcgid: can't apply process slot for /var/www/creativecommons.org/metadata_scraper/bin/scraper.fcgi" | 18:00 |
| nkinkade | Does anything immediately occur to either of you? | 18:00 |
| mralex | no idea | 18:00 |
| nyergler | JED3, googling seems to imply that it's a permissions issue | 18:02 |
| nyergler | http://www.megalinux.net/mod_fcgid-cant-apply-process-slot/ | 18:02 |
| nyergler | http://www.docunext.com/wiki/Apache_FastCGI#mod_fcgid:_can.27t_apply_process_slot_for... | 18:02 |
| *** jgay has quit IRC | 18:03 | |
| *** luisv has quit IRC | 18:05 | |
| *** luisv has joined #cc | 18:06 | |
| *** Kaetemi has quit IRC | 18:08 | |
| *** Kaetemi has joined #cc | 18:08 | |
| JED3 | nyergler: nkinkade figured it out | 18:17 |
| nkinkade | What was it? | 18:17 |
| JED3 | suexec won't run anything thats group writable | 18:17 |
| nkinkade | Ah. | 18:17 |
| nkinkade | That's a good one to tuck into the back of my head. | 18:18 |
| JED3 | :/ good to know in the future before i go down that rabbit hole again | 18:18 |
| nyergler | didn't we just talk about that a week ago? | 18:18 |
| nyergler | or maybe that was with webber | 18:18 |
| JED3 | yeah i'd never have dealt with that before | 18:18 |
| JED3 | nkinkade: i still need to change the logrotate, when do the logs normally rotate? | 18:19 |
| nkinkade | JED3: I'm not sure of the exact time. | 18:20 |
| nkinkade | I'm happy to do it, or you can look in /etc/logrotate.d | 18:20 |
| JED3 | was going to pack up and head into the office as my battery is dying | 18:20 |
| nkinkade | JED3: So I should just have logrotate set the logfile to owner www-data with mode 644? | 18:21 |
| nkinkade | Or 640, or evern 600? | 18:21 |
| JED3 | nkinkade: i just editted it to "john everette" | 18:21 |
| JED3 | everett* | 18:21 |
| nkinkade | JED3: everett needs to have write permissions. | 18:21 |
| JED3 | ok | 18:21 |
| JED3 | did you change it to 660 or was it already like that? | 18:22 |
| nkinkade | Not only to the log file, but also for the directory. | 18:22 |
| JED3 | nkinkade: everett has write | 18:22 |
| nkinkade | Cool. And those perms are already set in the logrotate conf file, right? | 18:23 |
| JED3 | does www-data need write, is that what you meant? | 18:23 |
| JED3 | yes | 18:23 |
| JED3 | create 660 john everett | 18:23 |
| *** oshani has joined #cc | 18:23 | |
| JED3 | heading into the office now, back in a bit | 18:24 |
| nkinkade | But JED3: should the script have create 660 www-data everett? | 18:25 |
| JED3 | nkinkade: no, i ended up reverting the suexec user back to john while i was trying to figure out what was going on | 18:26 |
| *** Imz has joined #cc | 18:26 | |
| nkinkade | Ah, okay, we'll now you know where to change it if you decide to use www-data at some point. | 18:26 |
| JED3 | i suppose i could change it to www-data if it makes the most sense for us to do it that way | 18:26 |
| JED3 | yeah | 18:27 |
| nkinkade | I don't think it really matters. john is fine. | 18:27 |
| JED3 | cool | 18:27 |
| *** JED3 has quit IRC | 18:32 | |
| *** TDJACR has quit IRC | 18:42 | |
| *** TDJACR has joined #cc | 18:42 | |
| *** Imz has left #cc | 18:43 | |
| *** TDJACR has quit IRC | 18:46 | |
| *** TDJACR has joined #cc | 18:47 | |
| nyergler | nkinkade, are you looking at the wiki? | 18:59 |
| nkinkade | nyergler: It seems to be up to me. | 18:59 |
| nkinkade | Is it down for you? | 18:59 |
| nyergler | http://wiki.creativecommons.org/Grants | 18:59 |
| nkinkade | That page loads fine for me. | 18:59 |
| nyergler | i get: | 19:00 |
| nyergler | Fatal error: Allowed memory size of 52428800 bytes exhausted (tried to allocate 74 bytes) in /var/www/wiki.creativecommons.org/www/extensions/AbuseFilter/AbuseFilter.i18n.php on line 5096 | 19:00 |
| nyergler | (for main page) | 19:00 |
| nyergler | Fatal error: Allowed memory size of 52428800 bytes exhausted (tried to allocate 1966080 bytes) in /var/www/wiki.creativecommons.org/www/extensions/SemanticForms/languages/SF_Messages.php on line 4637 | 19:00 |
| nyergler | for Grants | 19:00 |
| nyergler | also, nagios sent an email with "String not found" | 19:00 |
| nkinkade | Interesting. I'm not seeing that for some reason, even upon refresh. | 19:00 |
| nyergler | nkinkade, seems to be fine now | 19:02 |
| nyergler | weird | 19:02 |
| nkinkade | nyergler: One thing is that had APC disabled for testing. Perhaps that was causing an issue? | 19:05 |
| nkinkade | But strangely it wasn't causing an issue for me here?? | 19:06 |
| nkinkade | nyergler: For what it's worth, I think I may have found the problem, and it was Jesse W. that inadvertently pointed it out. Special:NewFiles, hangs, to the point where Apache never even logs the requests. | 19:09 |
| nkinkade | While I was busy trying to figure out why, the machine locked up and I had to reboot. | 19:09 |
| nkinkade | I wasn't sure if it was my testing that did it, or that in combo. with some crawlers. | 19:09 |
| nkinkade | Sadly wiki-staging, doesn't exhibit the problem, so I can't do testing on the testing site. | 19:10 |
| nyergler | sigh | 19:11 |
| *** JED3 has joined #cc | 19:12 | |
| JED3 | nyergler: have a sec? | 19:22 |
| nyergler | JED3, sure | 19:22 |
| greg-g | JED3: btw, did you get a chance to take a look at the Humanizer's output for More Permissions when using http://grossmeier.net/files/test.html ? | 19:23 |
| greg-g | should I just open a ticket instead of bugging you again? ;) | 19:23 |
| JED3 | greg-g: no, actually you raised an interesting question for nry and I | 19:27 |
| greg-g | uh oh | 19:27 |
| greg-g | my bad | 19:27 |
| greg-g | :) | 19:27 |
| JED3 | we didn't think about the case where people may be using mailto links for cc+ | 19:28 |
| greg-g | ahh, yeah | 19:29 |
| *** luisv has quit IRC | 19:33 | |
| *** luisv has joined #cc | 19:35 | |
| *** luisv has quit IRC | 19:38 | |
| *** luisv has joined #cc | 19:45 | |
| *** luisv has quit IRC | 19:50 | |
| *** michi_ has joined #cc | 19:57 | |
| nkinkade | mralex: Do you know if there is any reason we should be using ImageMagick to convert images on the wiki as opposed to PHPs internal tools? | 20:19 |
| mralex | nkinkade: do we have a version of php installed that has support for its built in image handling features? (libgd, if i recall) | 20:21 |
| nkinkade | mralex: We do: http://wiki.creativecommons.org/phpinfo.php | 20:22 |
| mralex | nkinkade: then i see no reason not to use it — assuming it's just a configuration option away? | 20:23 |
| nkinkade | mralex: It is, and the default is to use gd, but for some reason we had our install using ImageMagick. | 20:23 |
| nkinkade | Turns out it was ImageMagick dieing causing Special:NewFiles to hang. | 20:24 |
| mralex | ahh | 20:24 |
| nkinkade | I think ImageMagick was perhaps hittings a shell memory limit. | 20:24 |
| mralex | i imagine there was a reason for it, maybe we didn't have GD installed. | 20:24 |
| nkinkade | Yup. I re-enabled ImageMagick and bumped $wgMaxShellMemory to a much higher value and it worked. | 20:26 |
| nkinkade | But since we don't need to do that tweaking, I just set the wiki to use GD. | 20:27 |
| *** wormsxulla has quit IRC | 20:32 | |
| mralex | nkinkade: you know, the only reason i can think of why we weren't using GD was php's own memory limits... | 20:40 |
| greg-g | JED3: awesome, thanks for updating the humanizer | 20:41 |
| nkinkade | mralex: Hmm. I guess let's just see how things go without ImageMagick. | 20:42 |
| JED3 | greg-g: thats actually the code for the deeds :) | 20:44 |
| greg-g | oh | 20:44 |
| greg-g | oh right, because your humanizer just uses that? cool | 20:44 |
| JED3 | yep | 20:44 |
| greg-g | neat-o | 20:45 |
| JED3 | greg-g: http://code.creativecommons.org/~john/?url=http://grossmeier.net/files/test.html | 20:45 |
| greg-g | thankya much | 20:47 |
| JED3 | thank you | 20:47 |
| greg-g | any time | 20:48 |
| *** wormsxulla has joined #cc | 20:49 | |
| *** oshani has quit IRC | 20:57 | |
| *** oshani has joined #cc | 20:58 | |
| *** michi_ has quit IRC | 21:19 | |
| *** tantek has joined #cc | 21:21 | |
| *** sama_ has quit IRC | 21:55 | |
| *** oshani has quit IRC | 21:55 | |
| mralex | nkinkade: um, it appears the source= querystring isn't overriding the default contribution source for OneClick | 22:11 |
| mralex | which i thought i had overridden temporarly in OneClickConfig.php, but apparently that file isn't used any more. :/ | 22:12 |
| nkinkade | mralex: Hmm. You're right. Let me check it. | 22:12 |
| mralex | i see it's defined in OneClick.class | 22:12 |
| nkinkade | mralex: The only file used now is OneClick.class.php (and a small wrapper/router OneClick.php) | 22:13 |
| nkinkade | mralex: Do you want me to fix it, or are you in there doing it? | 22:14 |
| nkinkade | I see the issues. | 22:14 |
| nkinkade | If you're not already doing it, then I'll fix it now. | 22:14 |
| mralex | i'm peeking at the code, but you know where to go, so i'll let you handle it :) | 22:15 |
| nkinkade | mralex: in __construct I was referencing $this->_request_data['source'], which didn't exist. | 22:16 |
| nkinkade | Should have been $this->_request['source'] | 22:16 |
| nkinkade | I apparently change the name, but failed to update that one. | 22:16 |
| mralex | ah, i see | 22:16 |
| nkinkade | I'll go in and manually change the source for any contributions since ... ??? | 22:16 |
| nkinkade | What date? | 22:16 |
| nkinkade | mralex: ^^ | 22:17 |
| mralex | since june 1 that haven't already been modified | 22:17 |
| *** oshani has joined #cc | 23:00 | |
| JED3 | nyergler: ping | 23:02 |
| nyergler | JED3, pong | 23:02 |
| JED3 | nyergler: going to ask about how to handle restricting openid's in ccnet.php, but i'd rather do it next week | 23:04 |
| nyergler | JED3, ok | 23:04 |
| *** akozak has quit IRC | 23:13 | |
| *** tvol has quit IRC | 23:14 | |
| nyergler | nkinkade, we were thinking hardware RAID for that server, right? | 23:28 |
| nkinkade | nyergler: We certainly can. | 23:28 |
| nkinkade | I don't think it's critical, but it's always nice. | 23:29 |
| nyergler | right | 23:29 |
| nyergler | ok | 23:29 |
| nyergler | just doing some poking at that | 23:29 |
| nkinkade | I think it doesn't cost anything else from Dell on those servers to get hw RAID, or at least not much more. | 23:29 |
| nyergler | nkinkade, well it seems like there's some incremental cost on the one i was looking at, and they also want you to get all the drives in the configuration, too | 23:30 |
| nyergler | :/ | 23:30 |
| nkinkade | nyergler: I think I had a decent box spec'ed a PowerEdge R410??? | 23:30 |
| paulproteus | Hardware RAID is usually worse than software RAID. | 23:31 |
| nkinkade | paulproteus: In what sense? | 23:31 |
| paulproteus | Most "hardware raid" is what Linux calls "fakeraid" which iirc relies on a Windows driver to do the RAID. | 23:33 |
| paulproteus | c.f. http://www.brentnorris.net/blog/archives/158 | 23:33 |
| paulproteus | And then there's the whole write-back cache thing that took down LiveJournal. | 23:33 |
| paulproteus | For 36 hours or something. | 23:33 |
| paulproteus | http://lj-dev.livejournal.com/670215.html | 23:33 |
| paulproteus | "But unknown to us, the raid cards didn't disable the write caching on the drives themselves.... which is frickin' useless!" | 23:34 |
| paulproteus | Read that paragraph for more. | 23:34 |
| paulproteus | The more incoherent caches you have, the more bad news you'll have in case of emergency. | 23:34 |
| paulproteus | Linux software RAID is smart about these things. I'd actually trust it. | 23:34 |
| paulproteus | Humorously www.usenix.org/event/lisa04/tech/talks/livejournal.pdf mentions that they like "Battery-backed write-back caches." | 23:35 |
| paulproteus | Sure, in 2004 they did. | 23:35 |
| nyergler | paulproteus, interesting; the adaptec story isn't that concerning to me... i'd be clear enough to get something reported to work with debian/ubuntu for hardware raid | 23:36 |
| nyergler | (having been around the host raid block myself a few times) | 23:36 |
| * paulproteus nods and shrugs. | 23:36 | |
| nyergler | but i am interested to hear that you are comfortable with software RAID | 23:36 |
| paulproteus | I wouldn't bother with hardware RAID. As far as I can tell, it buys you possible LiveJournal style insanity at the benefit of... nothing. | 23:36 |
| nyergler | i think the benefit is supposed to be that it offloads parity calculation from the CPU | 23:37 |
| paulproteus | Yeah, Linux md has been through years and years of testing. | 23:37 |
| paulproteus | Boo frickin' hoo. | 23:37 |
| paulproteus | You're going to use RAID1, right? | 23:37 |
| nyergler | 5 | 23:37 |
| * paulproteus rolls his eyes. | 23:37 | |
| paulproteus | I mean, *fine*... | 23:37 |
| paulproteus | I would look seriously at RAID1. | 23:37 |
| nyergler | for simplicity reasons? | 23:38 |
| paulproteus | Simplicity and performance. | 23:38 |
| nkinkade | paulproteus: How then would you handle 4 disks, for example? | 23:38 |
| paulproteus | Two RAID1s? | 23:38 |
| nkinkade | Right with a 50% loss of actual capacity. | 23:38 |
| nyergler | "disks are cheap", paulproteus says | 23:38 |
| nyergler | ;) | 23:38 |
| paulproteus | With blindingly fast 2TB disks costing $200-400, is that really a problem? | 23:38 |
| nyergler | nkinkade, raid 1 would actually make our life slightly simpler | 23:39 |
| paulproteus | Last I saw, the CC servers used, like, 100GB of storage each or something. | 23:39 |
| nkinkade | paulproteus: It would just mean a more expensive server with an enclosure for at least 8 disks. | 23:39 |
| nyergler | wrt to the Amz move | 23:39 |
| paulproteus | Huh? re: 8 disks | 23:39 |
| paulproteus | You need to store 8TB? | 23:39 |
| paulproteus | Plus the in-case-of-emergency scenario is so easy with RAID1. | 23:39 |
| nkinkade | Why does RAID1 make life simpler over RAID5? | 23:39 |
| paulproteus | "Oh, the first disk doesn't boot? Okay, fall over to the second." Every BIOS has that built-in. Plus resilver is faster. | 23:39 |
| nkinkade | It's all handled by either the software driver or the hardware. | 23:39 |
| nyergler | nkinkade, i guess i wasn't convinced you could start an array from a disk with data on it (non-destructively) | 23:40 |
| nyergler | if that's possible, nm my comment | 23:40 |
| paulproteus | (Honestly, eight disks?) | 23:40 |
| nkinkade | nyergler: Perhaps you can't. | 23:40 |
| nkinkade | paulproteus: No, but maybe 3. | 23:41 |
| nkinkade | It would be 6 with a RAID1 setup. | 23:41 |
| paulproteus | How many TB of storage do you need? | 23:41 |
| nyergler | nkinkade, what's our current backup usage look like? | 23:41 |
| paulproteus | Disk speed is usually the bottleneck on a server. That, and not enough RAM (and then swapping, God help you). | 23:41 |
| nkinkade | 4T may not be enough in the long run. | 23:41 |
| nkinkade | nyergler: Order what makes most sense to you. | 23:41 |
| nyergler | nkinkade, gee, thanks :) | 23:41 |
| paulproteus | RAID5 with writes is tragic compared to RAID1. | 23:41 |
| nyergler | paulproteus, that is true (although "tragic" is probably a little strong ;) ) | 23:42 |
| paulproteus | I don't mean to push you guys around, just wanted to share my thoughts. (-: | 23:42 |
| nyergler | btw, i learned about RAID 6 today | 23:42 |
| nyergler | who knew | 23:42 |
| paulproteus | Oh God. | 23:42 |
| nkinkade | This is just a log storage and backup machine. | 23:42 |
| nyergler | paulproteus, i appreciate the input | 23:42 |
| nkinkade | Speed doesn't really matter. | 23:42 |
| paulproteus | nkinkade: Oh, that makes a huge difference! (-: | 23:43 |
| paulproteus | Yeah, in that case, whatever. I almost wouldn't bother with RAID. RAID 5, by all means. | 23:43 |
| nkinkade | Reliability does matter, but I'm not hearing anyone say, "Oh, never, *ever* use RAID5 for backups, because your data will be lost surely." | 23:43 |
| paulproteus | There was that bad summer I had when a RAID1 disk failed, and within 36h the other RAID1 disk failed. | 23:43 |
| nyergler | RAID 6! ;) | 23:44 |
| nkinkade | 4TB is enough for the near future. But if we get a machine with only a 4 drive enclosure and 2 RAID1 pairs, then we have no room to grow. | 23:44 |
| paulproteus | I like 4-drive 1U machines. | 23:44 |
| paulproteus | A lot. | 23:45 |
| paulproteus | Or 12-drive 2U machines. | 23:45 |
| paulproteus | Man, I really want one of those. | 23:45 |
| nkinkade | paulproteus: That is what I had in mind. | 23:45 |
| paulproteus | I have a 4-drive 1U. That's nice, but 12 disks... man. | 23:45 |
| nkinkade | A 1U machine with 4 possible drives. | 23:45 |
| * paulproteus nods. | 23:45 | |
| nyergler | paulproteus, do you build your own machines or buy assembled w/o OS? | 23:45 |
| nkinkade | My main reason for thinking RAID5, was to maximize the space and storage capacity of that 1U machine. | 23:45 |
| paulproteus | Yeah, RAID 5 for backup storage makes good sense. Don't let me stop you; performance doesn't matter. | 23:46 |
| paulproteus | (in the way it would for many other uses) | 23:46 |
| paulproteus | (i.e., your disks/filesystem will never be the bottleneck) | 23:46 |
| nkinkade | Write-back caches and battery backed RAID cards don't seem like an issue for us. | 23:46 |
| paulproteus | nyergler: I build my own, but if I were you I'd try to find a SuperMicro-based assembler that would assemble+warranty a whole machine. | 23:47 |
| nkinkade | It was simply a matter of making the most of space (physical space, not just disc capacity). | 23:47 |
| paulproteus | The crazy awesome computer store on 4th st & Howard (iirc)... | 23:47 |
| paulproteus | ...Central Computer, right... | 23:47 |
| paulproteus | ...is one of those. | 23:47 |
| paulproteus | Building servers is something I think I'm done with. | 23:47 |
| nyergler | paulproteus, yeah, i'm done with it, too | 23:47 |
| nkinkade | nyergler: Seriously, if you think RAID1 is the way to go, then it's all the same to me. I think it'll mean paying more for a 2U server to be sure we have some growing room, but the budget is your concern, not mine. | 23:48 |
| nyergler | nkinkade, right | 23:49 |
| nyergler | ok, i'm not doing anything this afternoon, will return to this monday | 23:49 |
| paulproteus | Now that I understand, I'd go software RAID5. Sorry to push so hard then immediately about-face when I learned context. (-: | 23:49 |
| nyergler | :) | 23:49 |
| paulproteus | I guess hardware RAID5 is good because you can boot off a degraded array. | 23:50 |
| paulproteus | (i.e., you can put / on RAID5 that way) | 23:50 |
| nyergler | right | 23:50 |
| paulproteus | If you go that route, majorly majorly do your homework. | 23:51 |
| paulproteus | I feel like there's so many possible screw-ups with hardware RAID that I don't want to touch it. | 23:51 |
| nyergler | yeah | 23:51 |
| paulproteus | But there are surely people who know this stuff and could happily help. | 23:51 |
| paulproteus | Some of them might even hang out on #cernio (a Bay Area tech cooperative) | 23:51 |
| paulproteus | esp. since you're CC and all that (-: | 23:51 |
| nyergler | ok cool | 23:52 |
| paulproteus | P.S. I love that I can buy a toaster oven on Newegg. | 23:52 |
| nyergler | lol | 23:52 |
| nyergler | well i'm out for the day... ttyl guys :) | 23:52 |
| *** nyergler has quit IRC | 23:52 | |
| paulproteus | I'll go back to my regular plan of listening to Rilo Kiley and cleaning house. | 23:52 |
| paulproteus | nkinkade: Sorry to barge in and sound like a know-it-all when I clearly didn't (-: | 23:52 |
| *** JoiIto has quit IRC | 23:58 | |
Generated by irclog2html.py 2.6 by Marius Gedminas - find it at mg.pov.lt!