Back to Home / #xen / 2005 / 04 / Prev Day | Next Day
#xen IRC Logs for 2005-04-05

---Logopened Tue Apr 05 00:00:56 2005
00:05--- ---> matta-lt [~matta@] has joined #xen
00:35--- ---> Shaun [] has joined #xen
02:30--- <<-- anthill333 [] has quit (Remote host closed the connection)
03:02--- <<-- rusty [] has quit (Ping timeout: 480 seconds)
03:46--- ---> athomas [] has joined #xen
04:04mael Hi
04:51* mael went to
04:52mael very instructive
04:52--- <<-- Shaun [] has quit (Read error: Connection reset by peer)
05:00--- ---> jesse_ [] has joined #xen
05:00--- User: *** jesse_ is now known as wenchien
05:38knewt mael: omg!!!
05:38mael yeah exactlu
05:39mael -u+y
05:42knewt gah html only emails are hateful
05:42knewt would be nice if xen-users and xen-devel rejected them
05:44mael yeah well this is a bit extreme
05:45knewt what, rejecting them? i don't think so. 99.99999999999% of all html-only emails are spam after all
05:54--- ---> Beaky [] has joined #xen
06:02knewt surriel: still around?
06:13surriel yes
06:14knewt you run the changelog bot, right?
06:14surriel *nod*
06:15knewt any chance of it mentioning which repository each change is for?
06:15surriel good idea, I need to figure out how to do that ;)
06:17* surriel adds that to his todo list
06:45--- <<-- athomas [] has quit (Quit: Leaving)
08:09--- <<-- Method [] has quit (Ping timeout: 480 seconds)
08:13--- ---> Method [] has joined #xen
08:29--- <<-- monrad [] has quit (Quit: Leaving)
08:32--- <<-- DEac- [] has quit (Ping timeout: 480 seconds)
08:33--- ---> athomas [] has joined #xen
08:36--- <<-- muligone [~muli@] has quit (Quit: My damn controlling terminal disappeared!)
08:45--- ---> DEac- [] has joined #xen
08:49--- ---> muli_ [] has joined #xen
08:53--- <<-- plars [] has quit (Quit: ran screaming into the woods)
08:55--- User: *** unriel is now known as riel
09:23--- ---> hollis [] has joined #xen
09:25--- ---> rharper [] has joined #xen
09:30--- ---> plars [] has joined #xen
09:47--- <<-- jeroney [] has quit (Remote host closed the connection)
10:30--- ---> homebaum [] has joined #xen
10:59--- <<-- Beaky [] has quit (Remote host closed the connection)
11:05--- ---> jeroney [] has joined #xen
11:10jeroney surriel: Did you try compiling x86-64 xen linux yesterday?
11:11--- <<-- JViz [] has quit (Ping timeout: 480 seconds)
11:19--- <<-- hollis [] has quit (Quit: leaving)
11:36--- ---> hollis [] has joined #xen
11:40--- ---> niv [] has joined #xen
11:47--- ---> niv_ [] has joined #xen
11:49--- <<-- niv [] has quit (Ping timeout: 480 seconds)
11:51--- <--- niv_ [] has left #xen ()
11:54--- ---> niv [] has joined #xen
11:56--- <<-- athomas [] has quit (Quit: Leaving)
12:00--- ---> cbrown [] has joined #xen
12:03hollis riel: is there a way to tell from the bk commit mails which tree the changeset is for?
12:09riel not yet
12:09riel jeroney: nope
12:10hollis riel: also:
12:10hollis Reply-To: Xen Development List <>
12:10hollis To:
12:10hollis I guess the sourceforge things needs updating
12:10riel whooops
12:10riel I'll do that right now
12:11riel thanks
12:11hollis sure
12:12riel actually, take a look at the X-BK-Repository: header
12:12hollis ah, hmm
12:13hollis nice, I just added that to Evolution's header list
12:17knewt ooh, nice. and the changeset key
12:17hollis ohhh, things got backed up
12:18hollis this mail is from 23 March
12:18hollis no wonder I was confused
12:18knewt and added to my .muttrc
12:19riel VFS: Cannot open root device "sda1" or unknown-block(0,0)
12:19* riel wonders ...
12:23riel xen_blk: Initialising virtual block device driver
12:23riel Could not probe disks (0)
12:25tab hm ?
12:25tab riel: is that new ?
12:25riel it's a very straightforward config file, too
12:25riel no idea why it's not working
12:26tab could not probe disk is not very good
12:26riel disks = [ 'file:/root/perf2.img,sda1,w' ]
12:26riel oh duh
12:26riel disk =
12:26* riel found a colleague's problem ;)
12:27* tab almost had a heart attack ;)
12:30* riel fixes the missing device nodes, too
12:30riel guess I'll let him figure out the missing /etc/inittab himself ;)
12:40--- ---> monrad [] has joined #xen
12:42cw riel: actually, a spam perl-script or whatever as a 'checker' wouldn't be a bad thing
12:56-!- Irssi: #xen: Total of 39 nicks 1 ops, 0 halfops, 0 voices, 38 normal
13:00--- <<-- cbrown [] has quit (Quit: Leaving)
13:15--- ---> MarkW [] has joined #xen
13:21--- ---> anthill333 [] has joined #xen
13:27--- <<-- jeroney [] has quit (Quit: Leaving)
13:43--- <--- anthill333 [] has left #xen ()
13:47--- ---> Shaun [] has joined #xen
13:59--- ---> jeroney [] has joined #xen
14:08* hollis requests remote cluebatting
14:14--- <--- niv [] has left #xen ()
14:25* riel remotely cluebats hollis ;)
14:27hollis :) if only that were all that's needed...
14:34plars hollis: I could walk down the hall and do it in person if you'd like? :)
14:35hollis I appreciate the offer, really
14:36rharper heh
14:37plars I'm here to help
14:37--- ---> tierra [] has joined #xen
14:47--- ---> JViz [] has joined #xen
15:02--- User: *** riel is now known as unriel
15:04--- ---> anthill333 [] has joined #xen
15:05--- <--- anthill333 [] has left #xen ()
15:08--- ---> niv [] has joined #xen
15:34--- <<-- matta-lt [~matta@] has quit (Quit: Hey! Where'd my controlling terminal go?)
16:09--- ---> MadMethod [] has joined #xen
16:10--- <<-- Method [] has quit (Ping timeout: 480 seconds)
16:17--- ---> Nigelenki [] has joined #xen
16:19--- <<-- Bluefox [] has quit (Read error: Operation timed out)
16:22--- <<-- MadMethod [] has quit (Ping timeout: 480 seconds)
16:31--- ---> athomas [] has joined #xen
16:49--- <<-- Nigelenki [] has quit (Read error: Connection reset by peer)
16:50--- ---> Nigelenki [] has joined #xen
16:55--- <<-- hollis [] has quit (Quit: leaving)
17:01--- <<-- athomas [] has quit (Quit: Leaving)
17:27--- ---> hollis [] has joined #xen
17:28--- <<-- Nigelenki [] has quit (Read error: Operation timed out)
17:31--- ---> Nigelenki [] has joined #xen
17:56--- ---> Method [] has joined #xen
18:00--- <<-- MarkW [] has quit (Remote host closed the connection)
18:27--- ---> MarkW [] has joined #xen
18:59--- <<-- rharper [] has quit (Quit: Leaving)
19:04--- <<-- hollis [] has quit (Quit: leaving)
19:21--- <<-- niv [] has quit (Ping timeout: 480 seconds)
19:36--- ---> rene_ [] has joined #xen
19:37rene_ with user mode linux, i was able to use hostfs to mount a filesystem
19:37rene_ is there an alternative for xen?
19:37rene_ my problem is that i need to preconfigure an image with root password, network config, hostname, etc, etc
19:38rene_ i was able to do this with hostfs, thanks to mounting the filesystem then writing code that grabs a config file from the hostfs mount and configuring the server during the rcS scripts
19:39--- ---> drbyte [] has joined #xen
19:40rene_ i was thinking about doing this through dhcp3, but im still stuck with the problem of root password and other configuration details
19:40caker rene_: if you can, loop mount it on the host and make your edits, before booting it
19:40rene_ caker: i was thinking that, and thats probably what im going to have to do
19:41rene_ doing the configuration through dhcp3 is fragile
19:42rene_ getting the configuration details over to the xen machine through scp or http is insecure
19:57MarkW rene_: for now, you should probably loopback mount the disk image
19:57MarkW why is it not secure to use scp?
19:58MarkW in the longer term, my xenfs project may be suitable for use in place of hostfs but it's a rather more complicated filesystem and so will take a while to complete
20:00rene_ MarkW: because the xm will then need to scp to dom0
20:00rene_ i was thinking about dom0 scping the config file to the xm, but what would trigger the dom0 to do that?
20:00rene_ it starts to get messy then
20:00rene_ hostfs was good for this sort of stuff
20:01MarkW ^ why not write a script in dom0 that does this?
20:02rene_ does what? scp configs over to the xm's?
20:02MarkW or alternatively, as script that loopback mounts and copies in the files, then starts the domain
20:02MarkW yes
20:02rene_ i could, but what would trigger the scp?
20:02rene_ and that also means that the xm's network configuration needs to be setup
20:03rene_ which at preconfiguration time, it isnt
20:03rene_ MarkW: loopback mount is good. im going to go down that path for now
20:03MarkW well, yes. if you configure the network in the domain config it'd work though.
20:03--- <<-- homebaum [] has quit (Remote host closed the connection)
20:03MarkW loopback mount is probably better and straightforwardly scriptable.
20:04* rene_ agrees
20:04MarkW mount -o loop should allocate a loop device automatically (although I'm not sure how you find out which)
20:05MarkW or you can grab the script Xend uses to allocate loop devices. that should make it easy to automate.
20:05mikegrb just mount /some/file.fs /mnt -o loop
20:05MarkW actually, i guess you probably don't need to care about which loop device it is - I guess it'll get unbound on umount
20:06MarkW so ignore what i just said and do what you were going to do anyhow ;-)
20:10mikegrb MarkW: ;)
20:10mikegrb MarkW: <3
20:11MarkW mikegrb: heh :-) in my defence I am rather sleep deprived... should get some shut-eye soon before the xen summit starts
20:13mikegrb certainly wouldn't hurt
20:14MarkW mikegrb: are you on any of the mailing lists? i don't recall seeing your name recently...
20:14mikegrb I haven't posted
20:14mikegrb I just subscribed today
20:14MarkW ah ok, cool. welcome :-)
20:15mikegrb MarkW: dunno if you remember Chris Aker's post about a leak in xend, it kind of got brushed off by ian
20:16MarkW yes, i remember
20:16mikegrb <-- check out this graph
20:16mikegrb growing by a couple hundred MB an hour
20:16mikegrb where it stops there at the end is when xm list quit being run to chack status
20:17* MarkW ponders the graph
20:18--- <<-- tierra [] has quit (Quit: Everybody wants to go to heaven, but nobody wants to die.)
20:19mikegrb so seems like some bit of code that runs in xend when xm list is called is living a data structure in the air so python thinks it is still in use and doesn't gc it
20:19MarkW help me out here - what does the scarily growing bit of the graph correspond to in the key? (i'm colour blind and there are too many colours there for me to make it out)
20:19mikegrb yes, the very top line above the filled in area is committed ram
20:19mikegrb the filled in area itself is swap
20:20MarkW it certainly doesn't look good.
20:20MarkW any thoughts why there's a little drop at the end?
20:20mikegrb no, I'm not too sure about that
20:20mikegrb I don't know if chris did anything
20:20MarkW and all you were doing was running xm list? how often?
20:21mikegrb very often ;)
20:21caker MarkW: it was a test -- watch -n0 xm list
20:21mikegrb caker: this is the graph from real world
20:21MarkW okie
20:21mikegrb did you have -n0 xm list the whole time?
20:21caker mikegrb: no, just for the last few hours or so
20:21mikegrb oh
20:21MarkW i guess there could be dangling references to some data structure that's stopping it getting garbage collected
20:22mikegrb the graph covers three days though
20:22MarkW errr, so it was growing hugely anyhow?
20:22mikegrb right
20:22mikegrb calculated it out, lemme find the numbers
20:22MarkW what load was it under? the domains producing lots of console output? lots of domain create / destroy / reboots?
20:23mikegrb 23:45:11 mikegrb | so 157.87 KB/min
20:23caker MarkW: Each domU has its console attached (for logging, etc). At first, I thought it was console output. After killing all the console sessions, memory still increased
20:24-!- Irssi: #xen: Total of 38 nicks 1 ops, 0 halfops, 0 voices, 37 normal
20:24caker that's when I started looking towards xm list (something else another process looks at each minute)
20:24caker MarkW == Mark Williamson ?
20:25mikegrb caker: that is what /wi says ;)
20:25MarkW i'd have thought in that case, doing more frequent xm lists would cause faster growth
20:25caker generally, not too manye reboots (one an hour, perhaps)
20:25MarkW yup, that's me ;-)
20:25caker heh
20:25MarkW but you didn't see this, right?
20:26MarkW it could be console, i guess. xend still has to process console info from the domain, even when you're not connected.
20:26MarkW hang on...
20:27--- <<-- drbyte [] has quit (Quit: Leaving)
20:29MarkW hmmm
20:30MarkW it's not doing anything obviously stupid (i.e. consoles are a circular buffer, so they don't get arbitrarily large)
20:30MarkW i guess there could be something wrong with the reference counting, that would do it
20:31MarkW (in C extensions you have to jiggle python ref counts directly, so there's potential scope for messing that up)
20:31mikegrb yeah, was going to ask about C extentions being used in that bit too
20:32--- ---> rusty [] has joined #xen
20:35MarkW mikegrb: can't see anything obvious there but i haven't looked very hard
20:36MarkW it'd be handy if you guys could keep an eye on this with a view to figuring out what induces it
20:36--- ---> drbyte [] has joined #xen
20:36MarkW in fact...
20:38MarkW ok, i'm just starting up my test box. i'll run xend overnight with no domains and see if it grows.
20:40MarkW then we can start to figure out what's different on yours. in the meantime, it'd be useful if you posted that graph to the list.
20:43MarkW gnight all, i need to get some sleep before the summit
20:43caker MarkW: thanks -- have fun :)
20:44MarkW will do. if Xend looks like it's going to break out of the machine and overwhelm the planet, you should let us know.
20:44--- <<-- DEac- [] has quit (Ping timeout: 480 seconds)
20:44caker heh
20:44caker we'll see how it performs since I rebooted this evening, too ..
20:45* MarkW goes...
20:45--- <<-- MarkW [] has quit (Quit: Kopete 0.10 :
20:56--- ---> DEac- [] has joined #xen
21:19--- <<-- rusty [] has quit (Remote host closed the connection)
21:20--- <<-- jeroney [] has quit (Quit: Leaving)
21:21--- <<-- rene_ [] has quit (Quit: [BX] "got bitchx?")
21:55--- ---> rusty [] has joined #xen
22:54--- <<-- Method [] has quit (Quit: Anac\xF8n\xD0a \xB7 "")
---Logclosed Wed Apr 06 00:00:30 2005