--- Log opened pon maj 24 00:00:49 2004 00:23:02>> netrose [netrose@SP2-24.207.228.55.charter-stl.com] has joined #vserver 00:23:09< Bertl> evening bobi! 00:23:27< netrose> Hi. 00:23:57< netrose> Bertl: I tried to use the "for loop" method to start the vservers faster... 00:24:05< Bertl> yep? 00:24:21< netrose> However, I think the total time required to start all the vservers with or without that method is about the same. 00:24:27< netrose> Do you agree with that? 00:24:39< Bertl> not necessarily ... 00:25:00< Bertl> a lot of time (in the startup scripts) is spent by waiting for something ... 00:25:06< netrose> Well, that's what it looks like on my vservers. 00:25:33< Bertl> could I see your loop statement please? 00:25:55< netrose> for n in ; do (vserver $n start &); sleep 2; done 00:26:14< Bertl> okay, how many vservers are there? 00:26:17< netrose> 17:26:03 up 16 min, 2 users, load average: 49.28, 30.73, 13.66 00:26:17< netrose> [ 00:26:22< netrose> about 30 00:26:43< Bertl> and this takes how long, until the last vserver did come up? 00:27:31< Bertl> s/did come/comes/ 00:29:21>> netrose_ [netrose@SP2-24.207.228.55.charter-stl.com] has joined #vserver 00:29:21>> netrose [netrose@SP2-24.207.228.55.charter-stl.com] has quit [Read error: Connection reset by peer] 00:29:47< netrose_> Sorry, got disconnected. 00:29:53< Bertl> np ... 00:30:25< netrose_> as I said, I am not timing this process, but I think it's not something observably faster. 00:30:46< Bertl> hmm, the load of 49 is from such a startup? 00:31:03< netrose_> right now... 00:31:04< netrose_> 17:30:53 up 21 min, 3 users, load average: 67.59, 52.26, 27.35 00:31:09< netrose_> yes such startup 00:31:17< Bertl> how many cpus are there? 00:31:22< Bertl> 1,2,4? 00:31:41< netrose_> I mean, httpd takes at least 5-7 seconds. That time has to pass. 00:31:45< netrose_> 2 CPUs 00:32:06< Bertl> okay, how long does a vserver startup take (a single one) in average? 00:32:22< netrose_> 30 seconds I guess. 00:32:53< Bertl> okay, so you can satisfy between 2 and 4 startups (considering the I/O stalls) 00:33:10< netrose_> Maybe, yes. 00:33:11< Bertl> let's make the sleep 2 a sleep 10 and try again ... 00:33:22< netrose_> What do you think that will achieve? 00:33:36< netrose_> I cannot try again because it's a production server. 00:34:18< Bertl> you are probably getting a slight speedup (with 2) on the startup, but you will lose plenty time at the end, where approx. 25 servers are starting simultaniously 00:34:25< netrose_> The server is quite responsive even during this process. 00:35:03< Bertl> so with 10 seconds sleep, you will reduce the number of simul servers ... and probably get a better satturation over time ... 00:35:26< netrose_> Anyway, I had another question 00:35:30< Bertl> yep? 00:36:11< netrose_> A process was using more than half of all available memory as shown in %MEM in ps -aux 00:36:33< netrose_> So he used all the swap too and crashed the server completely, so I had to reboot. 00:36:49< Bertl> okay .. 00:36:57< netrose_> Any ideas that could help? 00:37:16< netrose_> Besides writing a script that will check all processes each minute or so and just kill everything that is using more that 50% of the RAM? 00:37:46< Bertl> well, either using 1.9.x and memory limits or at least setting a per process limit, which is lower than, let's say 1/3 of the total memory should help 00:38:07< netrose_> per process limit with ulimit? 00:38:11< netrose_> for each vserver? 00:38:35< netrose_> 1.9.x is not stable, I assume? 00:38:40< Bertl> yep, that will prevent this .. but it might interfere with memory sensitive apps ... 00:38:53< Bertl> 1.9.x is development, 2.0.x will be stable ... 00:39:07< netrose_> Or, I can do that only for those vservers that are known offenders. 00:40:24< Bertl> sure, if you know that vserver xy is running mem hungry apps .. limiting that one would do the trick ... 00:40:41< Napalm> ok bert ive installed the old tools and got a vserver up, works fine 00:41:20< Bertl> if you are sure that no vserver is running a long time application, kill apps which take up 99% cpu over a longer time and/or use up too much memory would be an option too ... 00:41:34< Bertl> s/kill/killing/ 00:41:56< netrose_> That would have to be ascript to look at all processes 00:41:59< Bertl> you could check for that on a per hour basis or every 5 minutes ... 00:42:17< netrose_> That might be an unnecessary overhead. 00:42:22< netrose_> I don't know 00:42:48>> Doener_ [~doener@pD9E12423.dip.t-dialin.net] has joined #vserver 00:43:03< Bertl> maybe when there is enough interest (and time on my side) we could add some 'bad-app' spotting feature ... 00:43:13>> Doener_ is now known as Doener 00:43:22< netrose_> That would be very nice. 00:43:38< Bertl> something which looks at the behaviour of all processes from time to time and reports badly behaving processes somehow ... 00:44:00< Bertl> of course, the final decision has to be made in userspace on the host ... 00:44:05< netrose_> Sure. 00:49:46>> Doener` [~doener@pD9E12698.dip.t-dialin.net] has quit [Ping timeout: 480 seconds] 00:56:43>> circle- [~dude@24.137.95.180] has joined #vserver 00:56:56< Bertl> evening circle-! 00:58:05< circle-> hey :) 01:01:22< Bertl> what brings you here circle-? 01:01:41< circle-> trying to figure out what is v-serve 01:01:55< Bertl> wanna read some documentation about it? 01:02:01< circle-> sure 01:02:12< Bertl> http://vserver.13thfloor.at/Stuff/PAPER-05.4.txt 01:02:56< Bertl> (this is some meta format, and not the final version, but it should be readable) 01:04:58< circle-> So this is like a virtual server for all types of servers running under the control of one program? Did I get that right? 01:05:25< Bertl> scratch the control and the program, but basically yes 01:06:08< Bertl> think, like a bunch of real linux servers running on a single linux server 01:06:21< circle-> sounds like a great program if you can get the security down. 01:06:32< circle-> hack one get them all 01:06:49< Bertl> hmm, not really ... 01:07:16< Bertl> well, of course, if you hack the Host, you get the Guests for free ;) 01:07:40< circle-> hahaha thats what I meant 01:08:12< Bertl> of course, a good administrator will never expose the Host to the internet, which makes it a little harder, right? 01:08:27< circle-> true 01:09:21< Bertl> and things like tripwire can be installed on the host, watching over the Guest systems, undetected, which if done properly, increases the security for the Guests 01:09:23< circle-> now you would be looking at more then one peice of hardware 01:09:59< circle-> tripwire alone is not the best 01:10:17< Bertl> sure, that was just an example ;) 01:10:40< circle-> this sounds like a pretty good idea though 01:11:03< circle-> I may give it a try on one of my test boxes 01:13:26>> circle- [~dude@24.137.95.180] has quit [Quit: ] 01:19:30< Bertl> okay folks, EOD (End of Day) for me ... 01:19:41< Bertl> have a nice one, everyone, cya tomorrow ... 01:19:48>> Bertl is now known as Bertl_zZ 01:20:43< chaosle> n8 all 01:20:50>> chaosle [~yvan@port-212-202-168-55.dynamic.qsc.de] has left #vserver [Leaving] 01:28:33>> Napalm [~napalm@host81-7-20-161.adsl.v21.co.uk] has quit [Quit: ] 03:09:29< lilo> Hi Bertl_zZ 03:40:16>> deadguy [deadguy@bananajoe.big.du.se] has quit [Ping timeout: 480 seconds] 04:00:26>> infowolfe [~infowolfe@pcp04891550pcs.frnkmd01.md.comcast.net] has joined #vserver 04:00:53< Shotygun> evening wolfe 05:32:18>> deadguy [deadguy@bananajoe.big.du.se] has joined #vserver 05:42:23< infowolfe> hi 05:48:00>> shuri [~shushushu@cpu183.adsl.qc.bellglobal.com] has quit [Read error: Connection reset by peer] 06:35:55>> monrad [~monrad@213.83.190.229] has quit [Quit: Leaving] 08:12:36< eyck> hmm, brb. kernel testing. 08:12:39>> eyck [~eyck@62.233.189.138] has quit [Quit: leaving] 08:18:33>> eyck [~eyck@62.233.189.138] has joined #vserver 08:55:33>> Khahan [~Filbert@D57745B4.kabel.telenet.be] has quit [Remote host closed the connection] 09:00:02>> zev [~zev@masya.aviaserv.com.ua] has joined #vserver 09:00:28>> zev [~zev@masya.aviaserv.com.ua] has left #vserver [] 09:12:53>> Khahan [~Filbert@D57745B4.kabel.telenet.be] has joined #vserver 09:25:00>> Khahan [~Filbert@D57745B4.kabel.telenet.be] has quit [Read error: Connection reset by peer] 09:30:35>> virtuoso [~s0t0na@240ppp11.telegraph.spb.ru] has joined #vserver 09:31:06< virtuoso> Hiya 09:42:26>> Khahan [~Filbert@D57745B4.kabel.telenet.be] has joined #vserver 09:54:27>> JonB [~NoSuchUse@cpe.atm2-0-1031198.0x50a4ad0e.bynxx13.customer.tele.dk] has joined #vserver 09:54:36< JonB> Bertl_zZ: morning 09:57:22< virtuoso> JonB: Hi 09:57:37< JonB> hi virtuoso 09:58:53< virtuoso> How's it going? 09:59:16< JonB> busy at work, though looking for a danish company that sells linux vserver hosting 10:25:50>> virtuoso_ [~s0t0na@105ppp.telegraph.spb.ru] has joined #vserver 10:25:51>> virtuoso [~s0t0na@240ppp11.telegraph.spb.ru] has quit [Read error: Connection reset by peer] 11:41:53>> serving [~serving@213.186.189.12] has quit [Read error: Connection reset by peer] 12:21:12>> virtuoso [~s0t0na@213.158.7.174] has joined #vserver 12:21:12>> virtuoso_ [~s0t0na@105ppp.telegraph.spb.ru] has quit [Read error: Connection reset by peer] 12:46:47>> flock [~restless@l192-115-29-147.broadband.actcom.net.il] has joined #vserver 12:46:48>> anonymouscoward [~nwalsh@shaggy.internode.com.au] has quit [Remote host closed the connection] 12:46:56< flock> hey guys 13:29:54>> infowolfe [~infowolfe@pcp04891550pcs.frnkmd01.md.comcast.net] has quit [Read error: Connection reset by peer] 13:41:04>> serving [~serving@213.186.189.12] has joined #vserver 13:57:08>> mcp [~hightower@81.17.110.148] has quit [Ping timeout: 480 seconds] 13:58:50>> Bertl_zZ is now known as Bertl 13:59:18< Bertl> morning everyone! 13:59:31< flock> moring Bertl 13:59:58< albeiro> hi Bertl 14:00:00< virtuoso> Bertl: Hi 14:00:18< Bertl> hey flock, albeiro and virtuoso! 14:01:12< Bertl> what new with vserver? 14:01:19< Bertl> +s 14:01:49< flock> me? 14:02:26< Bertl> hey aren't you the development guys? ah, that's me, right! 8-) 14:02:45< flock> hehe 14:03:13< Bertl> I remember somebody was interested in improved/advanced networking and iptables stuff ... 14:03:34 * albeiro is interested in ipv6 :P 14:03:45< Bertl> ... and wanted to join development there .. any ideas who that was? 14:04:08< albeiro> i could grep log 14:04:15< Bertl> could you? 14:04:39< albeiro> yes, no problem 14:05:01< Bertl> so you ahve a good, mostly complete log then? 14:05:10< albeiro> waiting for courier mta to compile, so boring now 14:05:22< albeiro> yes, i have logs for all channels i am on 14:05:34>> Hest [~NoSuchUse@cpe.atm2-0-1031198.0x50a4ad0e.bynxx13.customer.tele.dk] has joined #vserver 14:05:47< Bertl> great! could you send this one to paul (sladen) or make it available for download (compressed) 14:05:57< Bertl> morning Hest! 14:05:58>> JonB [~NoSuchUse@cpe.atm2-0-1031198.0x50a4ad0e.bynxx13.customer.tele.dk] has quit [Read error: Connection reset by peer] 14:06:27< albeiro> sure, give me his email 14:06:55< Bertl> Paul Sladen 14:07:14>> mcp [~hightower@81.17.110.148] has joined #vserver 14:07:26< Bertl> morning mcp, how's life? 14:09:29>> Hest is now known as JonB 14:09:31< JonB> hey Bertl 14:09:41< JonB> Bertl: i was thinking about your remote reset... 14:09:53< Bertl> yep? what was the result? 14:10:05< JonB> Bertl: if i use it on a machine running the program from the same machine 14:10:30< JonB> will it reset or hang with power on the pin on the parport ?