Back to Home / #xen / 2005 / 04 / Prev Day | Next Day
#xen IRC Logs for 2005-04-19

---Logopened Tue Apr 19 00:00:30 2005
00:08--- ---> rusty [] has joined #xen
00:27--- <<-- knewt [] has quit (Ping timeout: 480 seconds)
00:35--- ---> knewt [] has joined #xen
01:18--- <<-- monrad [] has quit (Read error: Operation timed out)
02:50--- <<-- deac [] has quit (Ping timeout: 480 seconds)
02:59--- <<-- rusty [] has quit (Ping timeout: 480 seconds)
03:01--- ---> deac [] has joined #xen
03:05--- ---> hebutterworth [] has joined #xen
03:46--- ---> athomas [] has joined #xen
03:54--- ---> muli [] has joined #xen
04:03--- <<-- N [] has quit (Quit: (Session timeout))
04:15--- <<-- muli [] has quit (Ping timeout: 480 seconds)
05:32--- ---> soffi [] has joined #xen
06:46--- ---> Beaky [] has joined #xen
06:46--- <<-- rene- [] has quit (Quit: BitchX: often imitated, never duplicated!)
07:50--- ---> N [] has joined #xen
08:39--- User: *** unriel is now known as riel
08:43--- ---> hollis [] has joined #xen
08:44--- <<-- knewt [] has quit (Read error: Connection reset by peer)
08:45--- ---> knewt [] has joined #xen
09:04--- <<-- soffi [] has quit (Ping timeout: 480 seconds)
09:13--- <<-- matta-lt [~matta@] has quit (Quit: Hey! Where'd my controlling terminal go?)
09:15--- ---> gyver [] has joined #xen
09:17--- <<-- gyver [] has quit (Quit: )
09:19--- <<-- hollis [] has quit (Quit: leaving)
09:20--- ---> rharper [] has joined #xen
09:30--- <<-- knewt [] has quit (Ping timeout: 480 seconds)
09:41--- ---> hbaum [] has joined #xen
09:45--- ---> knewt [] has joined #xen
10:04--- ---> hollis [] has joined #xen
10:22--- ---> cfreak [] has joined #xen
10:28--- ---> Jd0gg [] has joined #xen
10:28--- <<-- Jdogg [] has quit (Read error: Connection reset by peer)
10:29--- User: *** Jd0gg is now known as Jdogg
10:44--- ---> itamarjp [] has joined #xen
10:48--- ---> caffeineOverload [] has joined #xen
10:50--- <--- caffeineOverload [] has left #xen ()
10:54--- <<-- Beaky [] has quit (Remote host closed the connection)
10:58--- ---> muli [] has joined #xen
11:13--- <<-- itamarjp [] has quit (Ping timeout: 480 seconds)
11:22--- <<-- knewt [] has quit (Read error: Connection reset by peer)
11:33--- ---> knewt [] has joined #xen
11:34--- <<-- nextime [] has quit (Quit: No windows for this server)
11:45--- ---> wmarvel [] has joined #xen
11:47knewt has acpi stuff been moved into dom0 in -unstable yet?
11:52--- <<-- knewt [] has quit (Quit: leaving)
11:58--- ---> knewt [] has joined #xen
12:15--- <<-- athomas [] has quit (Quit: Leaving)
13:25--- ---> mael_ [] has joined #xen
13:31--- <<-- wmarvel [] has quit (Quit: Leaving)
13:37--- ---> tierra [] has joined #xen
13:44--- <<-- Jdogg [] has quit (Read error: Connection reset by peer)
13:45--- ---> Jdogg [] has joined #xen
14:12--- <<-- rharper [] has quit (Read error: Connection reset by peer)
14:26--- ---> rharper [] has joined #xen
14:38--- ---> niv [] has joined #xen
14:44--- ---> cw [] has joined #xen
14:44--- Channel: mode/#xen [+o cw] by ChanServ
14:50--- <--- mael_ [] has left #xen (Fermeture du client)
15:09--- ---> lars||| [] has joined #xen
15:23--- <<-- cfreak [] has quit (Quit: .)
15:33--- <<-- lars||| [] has quit (Quit: using sirc version 2.211+KSIRC/1.3.11)
15:45--- ---> Shaun [] has joined #xen
15:53--- <<-- deac [] has quit (Ping timeout: 480 seconds)
16:04--- ---> deac [] has joined #xen
16:46--- ---> yarihm [] has joined #xen
17:08--- <<-- niv [] has quit (Quit: Quitting)
17:14--- ---> tierra|w_ [] has joined #xen
17:14--- <<-- tierra [] has quit (Read error: Connection reset by peer)
17:18--- User: *** riel is now known as unriel
17:24--- <<-- hollis [] has quit (Quit: leaving)
17:27--- <<-- hbaum [] has quit (Quit: Client exiting)
17:53--- <<-- yarihm [] has quit (Quit: Leaving)
17:57--- ---> homebaum [] has joined #xen
18:07--- <<-- N [] has quit (Quit: (Session timeout))
18:17--- <<-- lilo [] has quit (Read error: Operation timed out)
18:21--- <<-- rharper [] has quit (Quit: Leaving)
18:57--- <<-- Jdogg [] has quit (Ping timeout: 480 seconds)
18:57--- ---> itamarjp [] has joined #xen
19:36--- User: *** tierra|w_ is now known as tierra
19:37--- <<-- matta [~matta@] has quit (Quit: Sphinx may be image of a failed attempt to cross humans with lions.)
19:49--- ---> matta [~matta@] has joined #xen
19:49--- ---> lilo [] has joined #xen
19:56--- <<-- lilo [] has quit (Quit: bbiab)
20:01--- ---> lilo [] has joined #xen
20:17--- ---> anthony_ [] has joined #xen
20:17--- User: *** anthony_ is now known as aliguori_
20:18matta aliguori_: you actually here?
20:19aliguori_ matta: I am
20:20matta aliguori_: recommendation for vm-tools... right now it seems to only go by 'Domain-id' and 'id' for naming conventions
20:20matta note a lot of people specifically name their vm's
20:21matta so tasks such as 'vm list' should also print out the name given to the VM
20:21matta this is 0.0.3 i'm using here so sorry if that was already addressed
20:21aliguori_ matta: that's a great suggestion but a bit more involved then it would appear
20:21aliguori_ but it's being addressed :-)
20:21aliguori_ we're working on a very big update to vm-tools that should be coming out within the next couple of weeks
20:21matta [root@vm3 ~]# vm list | grep 20
20:21matta Domain-20 --b--- 31MB 32MB
20:21matta [root@vm3 ~]# xm li | grep 20
20:21matta jamesn 20 31 1 -b--- 138.4 9620
20:21aliguori_ indeed
20:22matta oh yeah?
20:22aliguori_ the hypervisor doesn't have any concept of names
20:22aliguori_ only IDs
20:22matta ok
20:22aliguori_ Xend maintains an ID=>name mapping
20:22matta so it's xend that is keeping track of the names
20:22matta gotcha
20:22matta ok, explains why it's not as trivial as it seems
20:22aliguori_ we don't because we're distributed.. they're no central daemon
20:22aliguori_ the update is creating a very minimal central daemon which will allow us to keep track of that sort of stuff
20:23matta is it written in C?
20:23aliguori_ yes
20:23matta awsome
20:23aliguori_ it will also replace xcs
20:23matta python is PITA
20:23matta which I assume is your drive behind the vm-tools...
20:23aliguori_ so it'll be an overall reduction of daemon code
20:23--- <<-- itamarjp [] has quit (Quit: )
20:23aliguori_ yes
20:23aliguori_ it's one of them
20:23matta another is easy direct API hooks?
20:24aliguori_ yes
20:24matta via libxen
20:24matta ok, i'll keep an eye out for the updates
20:24matta I run only one server with -unstable so I can test stuff like vm-tools
20:24matta but is enough to gain a feel for what is going on in unstable land
20:24aliguori_ we're actually looking to backport to 2.0.x in the very near future...
20:24matta yeah?
20:25aliguori_ yeah, i'm working on it as we speak
20:25aliguori_ :-)
20:25matta this will be part of the next release?
20:25matta which you say will be in the next week or two?
20:25aliguori_ it's a low-priority feature for the next week
20:25aliguori_ that's the plan
20:25aliguori_ next release*
20:25aliguori_ and yes, the plan is for another release in the next week or two
20:26--- ---> N [] has joined #xen
20:26aliguori_ we're also now using libxen instead of libxc.. which makes the error paths much much better
20:26matta cool, i'll keep an eye out... bye xend
20:26aliguori_ so the next release will be considerably more robust
20:26matta heh, I find it increasingly difficult to get meaningful errors from xend on some issues
20:26matta I sometimes experience the problem where it has the <name> as an id and then also Domain-id as id
20:26aliguori_ yes. it's not just meaningful errors, xend just plan ignores a number of error conditions
20:26aliguori_ yes
20:26matta so basically the same VM is using double the memory
20:27aliguori_ i do too
20:27matta I think that is due to xend
20:27matta not exactly yhe hypervisor
20:27matta is that what you think?
20:27aliguori_ vm-tools shows it properly though :-)
20:27aliguori_ when xend is doing that
20:27aliguori_ it must be caching results or something..
20:27matta hrm
20:27aliguori_ yes, it's xend, b/c i run xend and vm-tools in parallel and vm-list shows the right thing
20:27aliguori_ think*
20:27matta that explains why xm info shows memory usage change when I kill off id and then start up the domain again
20:28matta it's not actual memory being adjusted, but just what xend things is available
20:28matta er, thinks
20:28aliguori_ probably
20:28matta lol
20:28aliguori_ xend is a bit mysterous
20:28aliguori_ it's not very easy to groke through the code to figure out what its doing
20:28matta i'm still not sure, python is foreign to me
20:28matta I know a bit of C since it's what I started on
20:28aliguori_ i thought i was comfortable with python until i actually looked at xend
20:28matta but i'm a perl monkey :(
20:29aliguori_ :-)
20:29aliguori_ the multiple tools design of xend is specifically for perl and shell monkeys :-)
20:29matta at least with c programs I know how to use gdb and friends to find line numbers/files/etc
20:30aliguori_ yeah, well, my constant argument is that xend requires bindings to the hypervisor which take over 3,000 lines of C code to write.. vm-tools is under 2,000 lines of C
20:30matta nice
20:30aliguori_ so I cannot understand why python is even considered as an option
20:30matta and then the starting of xend when domains are running
20:30aliguori_ since it requires more C code than C...
20:30matta the xend memory growing issue
20:30matta vm-tools seems more and more attractive
20:31matta ok I gotta jet now... good to know vm-tools is making progress :)
20:32matta since next version will support 2.x i'm sure it will gain more testers
20:32aliguori_ matta: thanks :-)
20:32matta the backward compatibility with xm (via vm) is a nice touch
20:32matta makes switching little more than changing the command (or creating a symlink)
20:32matta anyhow, away for now
20:32matta later.
20:35aliguori_ cya
20:42--- <<-- alx_ [] has quit (Ping timeout: 480 seconds)
21:02caker <-- xend memory usage fun
21:02* surriel is very frustrated with the VDSO changes
21:02@surriel still haven't gotten xen0 to run without 'vdso=0'
21:05--- ---> aliguori- [] has joined #xen
21:05--- <<-- aliguori_ [] has quit (Read error: Connection reset by peer)
21:05aliguori- wow, i just crashed xen really hard
21:13--- ---> aliguori_ [] has joined #xen
21:13--- <<-- aliguori- [] has quit (Read error: Connection reset by peer)
21:13aliguori_ man, i seem to bring out the worst in xen
21:14--- <<-- aliguori_ [] has quit (Remote host closed the connection)
21:15--- <<-- homebaum [] has quit (Quit: Client exiting)
21:20--- ---> anthony [] has joined #xen
21:20--- User: *** anthony is now known as aliguori-
21:23aliguori- does anyone here have an opinion about using syslog in vm-tools?
21:24@surriel that may be useful in some situations
21:24@surriel but I'd keep it optional
21:24aliguori- and have the option to just store to a flat file?
21:24@surriel yes
21:24aliguori- yeah, that's not a bad idea..
21:25@surriel often it's easier to have a separate file per daemon
21:25aliguori- are there any syslog demultiplexers or something?
21:25@surriel especially when you want to debug something
21:25@surriel I don't know
21:25aliguori- gentoo comes with like three different syslog daemons..
21:27@surriel I consider that an argument in favor of per-daemon log files
21:27aliguori- yeah, i'm coding up the option right now.
21:27@surriel after all, if any of the syslog daemons did the right thing, there wouldn't be a need for the other two
21:27aliguori- syslog let's you do remote logging.. which is cool enough to support it
21:27@surriel absolutely
21:28* surriel would probably use syslog on production systems, flat files on development systems
21:29aliguori- yeah, i'll make it a command line option...
21:31* mikegrb commands surriel's line
21:31@surriel better yet, a config file option ;)
21:32aliguori- ewwww
21:32aliguori- no
21:32aliguori- :-)
21:32* aliguori- hates config files
21:32aliguori- besides, there's only logging in the long-running daemon
21:33@surriel oh ok, then /etc/sysconfig/xen works ;)
21:33aliguori- heh, you silly redhat guys with your sysconfig stuff :-)
21:34@surriel hehe
21:34aliguori- actually, i like it a lot.. that's why i want to keep everything command line..
21:34matta surriel:|ChangeSet@-7d
21:34matta that's for -testing, but might want to find the commit of it for -unstable and consider at least applying that patch to your FC4 RPM's
21:35* surriel looks
21:35@surriel I'll just upgrade to the latest nightly snapshot
21:35@surriel there's no reason to not do that
21:35matta yeah, commits have been slow lately
21:36matta nothing big, mainly bugfixes
21:36--- ---> Jdogg [] has joined #xen
21:36matta All domains terminated
21:36matta All domains terminated
21:36matta /etc/init.d/xendomains: line 38: log_success_msg: command not found
21:36matta hrm
21:37--- ---> hollis [] has joined #xen
21:38aliguori- matta: are you on a non-lsb compliant distro?
21:38matta FC3
21:38aliguori- log_success_msg is an lsb command.
21:38aliguori- oh, that's not right :-)
21:38matta ahhhh
21:38aliguori- it should be aliased in /lib/lsb/init-functions
21:39aliguori- perhaps you're missing a source /lib/lsb/init-functions?
21:42matta it exists
21:42matta i'll have to look at that later, minor error
21:43matta i'm still quite amazed at how well xen's QoS works
21:43matta even starting 50 VM's on boot dom0 is barely lagged
21:46aliguori- :-)
21:49matta surriel:|ChangeSet@-7d
21:52caker matta: which QoS are you talking about -- afaik, there's no disk QoS yet
21:54matta sure there is, didn't you read the white paper? it's in a primitive state
21:54matta unless it was removed in 2.0, which afaik it was not
21:55caker hmm .. are there any knobs to tweak, or just a default fairness type o deal
21:55aliguori- no
21:55caker in 2.0 and 2.1, I can starve other domains fairly easily
21:55matta from an early post
21:55aliguori- er, sorry, wrong window
21:55matta 4) I/O QoS - This is just round robin for now? I see it on the
21:55matta > roadmap. I have already envisioned the scenario with someone purchasing
21:55matta > a server with 64MB physical ram and adding a 512MB swap file inside
21:55matta > their server and just completely thrashing the disks. Is the current
21:55matta > system good enough to handle this (assuming some good SCSI disks are
21:55matta > being used) ?
21:56matta to which Mark W replied
21:56matta For network, you should be able to use any standard Linux QoS tools you want.
21:56matta For block IO, the requests are batched and serviced in a round robin fashion.
21:56matta This (obviously) doesn't provide service differentiation, or accounting. I
21:56matta think this is still on the roadmap and would certainly be cool to have.
21:56caker right
21:58matta well, it is better than nothing
21:59matta in my experience it does the job well
21:59matta just need to iron out the last little stability quirks of xen
21:59caker like what?
22:00matta I fix em or report em as I find em
22:01matta people need to take the plunge and put xen out in the open or else they won't be found, test suites will only do so much
22:23--- ---> itamarjp [] has joined #xen
22:23--- <<-- aliguori- [] has quit (Remote host closed the connection)
22:33--- <<-- itamarjp [] has quit (Quit: )
22:43--- <<-- JViz [] has quit (Ping timeout: 480 seconds)
22:59--- <--- VS_ChanLog [] has left #xen (Rotating Logs)
22:59--- ---> VS_ChanLog [] has joined #xen
23:19--- <<-- tierra [] has quit (Quit: Everybody wants to go to heaven, but nobody wants to die.)
23:37--- <<-- hollis [] has quit (Quit: leaving)
23:53--- ---> aliguori- [] has joined #xen
---Logclosed Wed Apr 20 00:00:03 2005