Back to Home / #linode / 2006 / 09 / Prev Day | Next Day
#linode IRC Logs for 2006-09-19

---Logopened Tue Sep 19 00:00:32 2006
00:13|-|emad [] has joined #linode
00:13<emad>hi, is there a general help channel here that can get me in contact with an ircop?
00:14|-|emad [] has left #linode []
00:16|-|coumbes [~coumbes@] has joined #linode
00:18|-|coumbes [~coumbes@] has quit []
00:41|-|Netsplit <-> quits: schultmc, JDM
00:42|-|Netsplit over, joins: JDM
00:42|-|Netsplit over, joins: schultmc
00:48|-|Netsplit <-> quits: schultmc, JDM
00:51|-|Dreamr_3 [] has joined #linode
00:56|-|Dreamr3 [] has quit [Ping timeout: 480 seconds]
00:59|-|Netsplit over, joins: JDM
00:59|-|Netsplit over, joins: schultmc
01:07|-|spr [] has quit [Quit: This computer has gone to sleep]
01:45|-|internat [] has quit [Quit: This computer has gone to sleep]
01:47|-|Electric1lf [] has joined #linode
01:54|-|ElectricElf [] has quit [Ping timeout: 480 seconds]
02:40|-|rko_ [] has joined #linode
02:45|-|rko [] has quit [Ping timeout: 480 seconds]
03:07|-|SpaceHobo [~spacehobo@] has joined #linode
03:09|-|Netsplit <-> quits: schultmc, JDM
03:16|-|Netsplit over, joins: JDM
03:16|-|Netsplit over, joins: schultmc
03:30|-|Netsplit <-> quits: schultmc, JDM
03:33|-|lucca [~lucca@] has quit [Remote host closed the connection]
03:34|-|Netsplit over, joins: JDM
03:34|-|Netsplit over, joins: schultmc
03:39|-|lucca [] has joined #linode
04:03|-|jblack [] has joined #linode
04:05<jblack>I'm spiking the virtual machine a bit as I onload the site, get things installed, etc.
04:07<jblack>I'm trying to be a good neighbor and all by spreading the tasks out. Is it there anything else that I can reasonably do? Does nicing tasks provide any benefit outside of the virtual machine?
04:44<JDM>jblack: because everyone has a commited mhz ratio, and the cpu divides equally (i think), it wouldn't be a problem if you didn't space the tasks
04:48|-|ophionman [] has joined #linode
04:48<ophionman>hey fella
04:49<ophionman>can someone help me with some weird network problems with a node in host8?
04:49<ophionman>it basically looks like the nic is down.
04:49<jblack>ok. thanks
04:50<ophionman>I can't get an dhcp ip and if I ifconfig the nic up, I can't ping the router.
04:50<ophionman>this happened after I rebooted my linode this morning
04:50<ophionman>bad.. bad call
04:56<@linbot>New news from forums: EHLO: different size allowed, from different servers to ext in Email/SMTP Related Forum <>
04:58|-|internat [] has joined #linode
05:37|-|ophionman [] has quit [Quit: ophionman]
06:03<@linbot>New news from forums: We sell Cheap Viagra ($1.41) and Cialis ($2.65) in General Discussion <>
06:03<SpaceHobo>linbot: TELL ME MORE!
06:03<SpaceHobo>he did.
06:03<SpaceHobo>12:03 -!- Irssi: Starting query in Linode with linbot
06:03<SpaceHobo>12:03 <linbot> SpaceHobo wants me to tell you: MORE!
06:43|-|linville [] has joined #linode
07:00|-|internat [] has quit [Quit: This computer has gone to sleep]
07:51|-|Antitribu [~simon@] has joined #linode
07:51|-|Antitribu [~simon@] has left #linode []
09:26|-|jungleg [] has joined #linode
09:57|-|spr [] has joined #linode
10:24|-|Pretender [] has joined #linode
10:24|-|Pretender changed nick to rslomkow
10:25<rslomkow>hello does anybody here know how to identify how much io a given userspace process is using on a linode 2.6 kernel ?
10:26<rslomkow>sorry 2.4 kernel
10:31<SpaceHobo>hey robin
10:31<SpaceHobo>I'd like to know how to do that in 2.6, myself
10:56|-|spr [] has quit [Quit: Spoon!]
11:01<rslomkow>well I found my IO demon, it was amavis-new
11:07<npmr>i stopped using amavis shortly after i started using it
11:08<npmr>the issue i had was memory usage leading to swapping
11:08<npmr>i can't see any good reason why a mail filter multiplexer should need several tens of megabytes of ram
11:13<rslomkow>I trimmed its ram usage, but its io usage was still insane
11:13<rslomkow>520 io/s (with it on) vs 30 io/s (with it off)
11:21<npmr>so that's actual reading and writing of files and not swap-in/swap-out?
11:27|-|SpaceHobo [~spacehobo@] has quit [Read error: Operation timed out]
11:42<rslomkow>npmr: no swap enabled
11:42<npmr>swapping does not require a swap partition
11:43<npmr>all pages of memory are included in the two-tier memory hierarchy
11:43<rslomkow>and that counts toward my system io?
11:43<npmr>any read from or write to your ubd* block devices hits your quota
11:44<rslomkow>vmstat didn't think I was doing much but iostat counted many writes, but I couldn't determine explictly what was doing it
11:44<npmr>that includes swap-in/swap-out of demand-paged memory mapped files
11:44<rslomkow>I suspect it must have been some of the virus definition files
11:44<npmr>yeah, vmstat does provides an excellend overview
11:45<npmr>what scanner are you using?
11:45<rslomkow>amavis-new and clamav
11:45<npmr>i had that same setup
11:45<rslomkow>no longer using either
11:45<npmr>scrapped it
11:45<npmr>i now just run clamscan from /etc/procmailrc on all delivered mail
11:46<rslomkow>I had to do a bit of tuning to reduce the memory footprint of the two, which is probably what made it start doing more io
11:46<npmr>it's a much smaller i/o hit, especially since you can avoid scanning relayed mail, or mail that's ultimately rejected
11:46<npmr>probably, yeah
11:47<rslomkow>good idea, maybe I should re-install clamscan
11:47<npmr>it's worked quite well for me
11:47<npmr>i don't know how my mail volume compares to yours though
11:48<npmr>in late august i did two things
11:49<npmr>i unsubscribed from lkml
11:49<npmr>and i started messing around with my postfix restrictions
11:49<npmr>most of that drop in the "delivered" statistic is accounted for in the lkml unsubscription
11:50<npmr>but anyway, with that drop in mail volume i went from roughly 40 ops/sec to roughly 20
11:51<npmr>the major components of the i/o hit were clamscan and bogofilter, but it wasn't unreasonable
11:51<npmr>amavis and clamav was unreasonable
11:51|-|spr [] has joined #linode
11:55<rslomkow>I receive around 5 thousand per day, about 1k get rejected per day from rbls, another 1k from rcpt and such, and my bogofilter does about 200-400 per day
11:55<rslomkow>then a few hundread per day actually delivered
11:55<rslomkow>the rest goes to other users
11:55<rslomkow>no that my io is back under control I should look at more monitoring software ;)
11:56<npmr>about half of the remaining 20 ops/sec is munin, i think
11:56[~]npmr shrugs
11:57<npmr>i think if i just had the data collection running and moved munin-update offsite that would be a big gain
11:57<npmr>but i lack the means anymore
12:10|-|SpaceHobo [] has joined #linode
12:18<@linbot>New news from forums: Reboot: host46 in System and Network Status <>
12:52|-|Dreamr3 [] has joined #linode
12:57|-|Dreamr_3 [] has quit [Ping timeout: 480 seconds]
13:03|-|jimcooncat [] has joined #linode
13:08|-|jimcooncat [] has quit []
13:34|-|jekil [] has joined #linode
13:56|-|jekil2 [] has joined #linode
13:56|-|jekil2 [] has quit []
14:26|-|jungleg [] has quit [Quit: CGI:IRC (Session timeout)]
14:28|-|eric_ [] has quit [Quit: BitchX-1.1-final -- just do it.]
14:43|-|jimcooncat [] has joined #linode
14:44|-|jimcooncat [] has quit []
14:49|-|ophionman [] has joined #linode
14:50<ophionman>hey fellas,
14:50<ophionman>can someone help me with my linode, I've put in a support request but thus far no luck
14:50<ophionman>basically my network is dead
14:50<ophionman>it can't send and receive aything out
14:50<@caker>ophionman: try rebooting?
14:50<rko_>what does ifconfig say?
14:50<ophionman>yeah tried that.
14:51<ophionman>even rebuilt the node
14:51<@caker>iptable rules?
14:51<ophionman>no iptables
14:51<ophionman>ifconfig says it is up
14:51<@caker>did you migrate recently?
14:51<ophionman>put can't ping router
14:51<ophionman>I'm scheduled to migrate
14:51<rko_>right IP, right mask?
14:51<ophionman>right ip right mask
14:52<ophionman>dhcp never gets an ip
14:52<@caker>ophionman: that's why then, you're IP won't work until you perform the migration -- your migration is to a new datacenter, thus requiring a new IP
14:52<ophionman>well it never gets a ip offer because it never reaches the dhcp server
14:52<ophionman>how do I trigger that?
14:52<@caker>ophionman: since you've rebuilt the node, if you delete the images the migration will take seconds
14:52<ophionman>I thought the migrations were on hold
14:52<@caker>ophionman: hold on...
14:53<ophionman>[root@localhost /]# dhclient eth0Internet Systems Consortium DHCP Client V3.0.1Copyright 2004 Internet Systems Consortium.All rights reserved.For info, please visit on LPF/eth0/fe:fd:42:dc:01:cbSending on LPF/eth0/fe:fd:42:dc:01:cbSending on Socket/fallbackDHCPDISCOVER on eth0 to port 67 interval 6DHCPDISCOVER on eth0 to port 67 interval 12DHCPDISCOVER on eth0 to 255.255.2
14:53<ophionman>boy that looks like shit
14:53<@caker>ophionman: what was your old IP?
14:53<@caker>ophionman: or, does it matter if you're given a new one?
14:53<@caker>that's the new one
14:54<@caker>lemme reassign a new-new IP, one sec :)
14:54<ophionman> is the old one
14:54<ophionman>according to dns
14:54<ophionman>caker, so how do I finish the migration?
14:54<@caker>ophionman: don't bother
14:54<@caker>ophionman: ok, reboot and you should have the old IP back
14:55<@caker>sorry about that -- that was an artifact from when we configurated your Xen migration
14:55<ophionman>I was going crazy.. now I have to spend the next hour setting things back up :(
14:56<ophionman>thank you tho.
14:56<ophionman>so, now I"m back on host8?
14:56<@caker>still on host8, yes
14:57<ophionman>all right fellas, later
14:58|-|ophionman [] has quit [Quit: CGI:IRC]
14:59<jblack>If the's a quote page, then the viagra conversation from 9 hours ago belongs on it. :0
15:28<Battousai>i demand a link
15:29<@caker>sb end
15:29<@caker>damn irssi :)
15:32<jblack>07:03 <@linbot> New news from forums: We sell Cheap Viagra ($1.41) and Cialis ($2.65) in General Discussion <>
15:32<jblack>07:03 < SpaceHobo> linbot: TELL ME MORE!
15:32<jblack>07:03 < SpaceHobo> Oh.
15:32<jblack>07:03 < SpaceHobo> he did.
15:32<jblack>07:03 < SpaceHobo> 12:03 -!- Irssi: Starting query in Linode with linbot
15:32<jblack>07:03 < SpaceHobo> 12:03 <linbot> SpaceHobo wants me to tell you: MORE!
15:39<SpaceHobo>linbot: TELL jblack MORE!
15:39<@mikegrb>mmm cake
15:39<SpaceHobo>linbot: tell mikegrb cake
15:39|-|linville [] has quit [Quit: Leaving]
15:45<lucca>jblack: are you that guy from dalnet?
16:06<jblack>I'm the same person on dalnet, yes.
16:09<SpaceHobo>linbot: tell linbot tell linbot tell linbot
16:09<jblack>I finished moving my websites up and it looks like I'm hovering near 1.0 cpu. I might have to buy more cpu once I change the mx records for mail and spamassasisn starts doing its magic.
16:15<jblack>tell linbot tell jblack says hi
16:15<jblack>Nope. Not recursive. =)
16:17<Battousai>buy cpu?
16:19<jblack>Its not a directly purchasable option. Its indirectly purchasable by selecting more expensive plans with less contention
16:22<@mikegrb>mmm cake
16:22<brocktice>didn't know about that one
16:22<Battousai>that's because you were not yet great enough to see it
16:22<Battousai>but now you are
16:22<Battousai>so HAPPY
16:23<brocktice>Arr, it be like scientology
16:26<jblack>Ok. it looks like i need to speak to someone about my plan.
16:27<Battousai>filing a ticket is probably best
16:30|-|jekil [] has quit [Ping timeout: 480 seconds]
16:44|-|internat [] has joined #linode
16:46<jblack>wouldn't it be interesting if there were a cpu token market at linode?
16:47<jblack>darn. becuase I'd be willing to buy a bit of cpu time for other nodes.
16:47<Battousai>perhaps if it were some sort of dark color market
16:47<Battousai>such as black
16:48<brocktice>I think a storage market would be much more interesting
16:48<JDM>you can use up to 100% of the host cpu when other nodes aren't using it
16:48<jblack>Yeah. I'd be willing to trade half of my filesystem space for a doubling of my cpu.
16:51<jblack>jdm: I'm not sure that's a good spot for me to go in, to rely on other neighbors being noncontestive.
16:52<jblack>I could be fine for a while, only to wake up weeks later to find myself in permanant starvation.
16:52|-|internat [] has quit [Quit: This computer has gone to sleep]
16:54<@caker>jblack: I don't think I've ever seen anyone under CPU contention (besides the obvious SETI@home abuser, or the like)
16:54<kvandivo>you can abuse SETI@home?
16:54<@caker>jblack: typically you'll hit disk i/o and/or network (although doubtful) contentions way before CPU
16:55<kvandivo>(sorry.. proofreading proposal text and finding lots of phrases that can be misinterpreted)
16:55<jblack>I might be the first.
16:55<@caker>what's eating cpu in your node, then?
16:56<jblack>Oh, its not what I'm doing now, but going to do.
16:56<kvandivo>and that's a sekret
16:56<jblack>I want to switch the mailserver over. I use spamassassin, because my various email addresses are all over the place.
16:57<jblack>So when spamc starts filtering, its going to be performing bayesian analysis
16:57<@caker>also, that cpu beancounter thing is ... just there as a high water mark ... the cpu graphs are way more accurate
16:58<jblack>I'm looking at the graph.
16:58<@caker>also, it IS possible to have more than 100%, because UML is split on the host into two processes -- one for kernel calls and one that spins for userspace
16:58<jblack>Right now I'm bouncing under/over 1.0 on the graph, with an upward trend.
16:58<@caker>eventually, either with Xen or with SMP-UML, you'll have virtual CPUs inside the VM itself, so that's one possible "upgrade" path
16:59[~]caker looks
16:59<@caker>oh, that's 1%, I
16:59<@caker>I'd say you've got some headroom left :)
16:59<jblack>Oh, 1% of a cpu. :)
16:59<@caker>not loadavg
16:59<jblack>Ok. That seems much, much more sane.
16:59<kvandivo>the main thing you have to worry about with SA is ram... big time
17:00<kvandivo>cpu isn't going to play a role, most likely
17:00<@caker>ya -- someone once said to make it run under one thread only
17:00<jblack>I couldn't make the numbers add up. I've 8 minutes of cpu time in 1.5 days worked out to the provision
17:00<jblack>Um, that wasn't in english
17:01<jblack>I couldn't make the numbers add up as 8 minutes of cpu time in 1.5 days seemed rather like a tight squeeze
17:01<jblack>Can you cancel ticket 21924?
17:46|-|spr [] has quit [Quit: "Ah, the bounce has gone from his bungie"]
18:20|-|spr [] has joined #linode
18:22|-|jekil [] has joined #linode
18:43|-|jekil [] has quit [Ping timeout: 480 seconds]
19:13|-|internat [] has joined #linode
19:13|-|flatronf700B [~flatronf7@] has joined #linode
19:48|-|internat [] has quit [Quit: This computer has gone to sleep]
19:59|-|internat [~internat@] has joined #linode
20:24|-|rko_ [] has quit [Read error: Operation timed out]
20:37|-|Battousai [~bryan@2002:d8a4:1c9e:2:20e:35ff:febc:b116] has quit [Remote host closed the connection]
20:58|-|internat [~internat@] has quit [Quit: This computer has gone to sleep]
21:58|-|Battousai [] has joined #linode
22:11|-|Battousai [] has quit [Quit: Changing server...]
22:16|-|Battousai [~bryan@2002:d8a4:1c9e:2:20e:35ff:febc:b116] has joined #linode
22:59|-|VS_ChanLog [] has left #linode [Rotating Logs]
22:59|-|VS_ChanLog [] has joined #linode
23:18<encode>is there any way of logging who is port forwarding what when logging in via openssh?
23:22|-|internat [] has joined #linode
23:23|-|internat [] has quit []
---Logclosed Wed Sep 20 00:00:20 2006