--- Day changed

--- Log opened Thu Aug 28 00:00:02 2003
01:45 -!- pirlouit [~peter@duvel.drunkcoders.com] has quit [Ping timeout: 480 seconds]
01:49 -!- pirlouit [~peter@64.162.195.202] has joined #uml
03:24 -!- pirlouit [~peter@64.162.195.202] has quit [Quit: Time to frag!]
03:24 -!- pirlouit [~peter@duvel.drunkcoders.com] has joined #uml
03:39 -!- Decdec [~ircap75@80.34.115.245] has joined #uml
03:39 < Decdec> hi
03:39 < Decdec> I need help about creating my own file system
03:40 < Decdec> I have followed the steps at UML official web
03:41 < Decdec> host% dd if=/dev/zero of=new_filesystem seek=100 count=1 bs=1M
03:41 < Decdec> Add an argument like the following to the UML command line:
03:41 < Decdec> ubd4=new_filesystem
03:41 < Decdec> mkreiserfs /dev/ubd/4
03:41 < Decdec> mount /dev/ubd/4 /mnt
03:41 < Decdec> and now....?
03:42 < Decdec> what way I copy my wished fs to mnt?
03:42 < Decdec> dd, cp or how?
03:50 < green> what's "your wished fs"?
04:12 -!- Decdec [~ircap75@80.34.115.245] has quit [Quit: ..]
04:22 -!- Michael2 [~chatzilla@algedi.dur.ac.uk] has joined #uml
05:40 -!- pirlouit [~peter@duvel.drunkcoders.com] has quit [Quit: Time to frag!]
06:51 -!- NotTheDuck [~ducky@65.40.2.89] has joined #uml
06:57 -!- TheDuck [~ducky@65.41.143.23] has quit [Ping timeout: 490 seconds]
07:27 -!- TheDuck [~ducky@65.41.1.156] has joined #uml
07:33 -!- NotTheDuck [~ducky@65.40.2.89] has quit [Ping timeout: 490 seconds]
09:05 -!- pixelfairy [~pixel@64.81.36.140] has joined #uml
09:23 -!- jax-work [~computer_@148.100.209.96] has joined #uml
09:43 -!- Stuart [~stuart@myrddraal.demon.co.uk] has joined #uml
09:59 -!- NotTheDuck [~ducky@65.41.7.202] has joined #uml
10:03 -!- Stuart [~stuart@myrddraal.demon.co.uk] has quit [Quit: leaving]
10:05 -!- TheDuck [~ducky@65.41.1.156] has quit [Ping timeout: 490 seconds]
10:10 -!- NotTheDuck [~ducky@65.41.7.202] has quit [Ping timeout: 490 seconds]
10:35 -!- TheDuck [~ducky@65.41.159.11] has joined #uml
11:40 -!- ticallion [~ticallion@213.175.160.236] has joined #uml
11:45 -!- ticallion [~ticallion@213.175.160.236] has quit [Quit: brb]
12:51 -!- Michael2 [~chatzilla@algedi.dur.ac.uk] has quit []
13:17 -!- jdike [~jdike@12.6.236.65] has joined #uml
13:17 < jdike> hi guys
13:24 -!- jdike [~jdike@12.6.236.65] has quit [Quit: Leaving]
13:38 -!- pirlouit [~peter@duvel.drunkcoders.com] has joined #uml
13:51 -!- coryb [~cory@24.124.36.198] has quit [Quit: leaving]
14:40 -!- jdike [~jdike@12.6.236.73] has joined #uml
14:46 < green> Hi Jeff!
14:47 < jdike> Hey Oleg
14:47 < green> jdike: Have you seen this directio idea for ubd? I think it might be used as one of the solutions for unnecessary cache pollution
14:48 < jdike> green: I've considered that, but ubd mmap pretty much does the same thing
14:48 < jdike> green: and it allows for sharing between UMLs
14:49 < green> yes.
14:49 < green> But the mmap code is new and experimental one (for developers) and directio can be used by usual people just now. (of course we'd need to increase ubd sector size to page size for it to work on 2.4)
14:50 < jdike> green: True
14:50 < jdike> green: BTW, I've also done COW v3
14:51 < jdike> green: so that the alignment within the COW file is correct for mmap
14:51 < jdike> green: which wasn't true for V2
14:51 < green> ah. great
14:51 < green> btw, mmap + cow should be extra tricky, I guess
14:53 < jdike> green: it's not horrible
14:53 < green> copy on write of some sort?
14:53 < jdike> green: the worst bit is moving a mapping from the backing file to the COW file in response to a segfault due a write access
14:54 < green> probalby you can do it the other way
14:54 < jdike> green: copy the data to the COW file, update the bitmap, map in the COW file page
14:54 < jdike> green: what other way?
14:54 < green> you can do private mapping in case of cow device. and then change the backing file when somebody issues actual IO request
14:55 < green> or does waiting on page/buffer
14:56 < green> does it sound too tricky too?
15:10 < jdike> green: I don't understand that
15:10 < jdike> green: Reword it/
15:10 < jdike> ?
15:15 < green> well.
15:15 < green> you have two files. readonly copy and copy on write file
15:16 < jdike> green: I'm with you so far :-)
15:16 < green> you map readonly one as MAP_PRIVATE. then all changes to it won't be written back.
15:16 < jdike> green: ah
15:16 < green> then when somebody does wait on page/buffer (or just submits io - you decide) - you write the page to cow file, unmap changed page and map COW file as MAP_SHARED
15:17 < jdike> green: then actually do the write when the block layer says to do it
15:17 < green> yes
15:17 < jdike> green: that still requires flipping the mappings around
15:17 < jdike> green: just later
15:17 < green> yes. but you do not need to deal with signals and stuff liek that
15:18 < jdike> green: and you have some excess host memory consumed in the meantime
15:18 < green> still tricky, but possibly easier. but may be not
15:18 < jdike> green: true, but the segfault isn't really the problem
15:18 < green> unfortunally both approaces are SMP race prone
15:19 < green> hm. actually signal approach can be handled easier
15:22 -!- jax-work [~computer_@148.100.209.96] has quit [Read error: Connection reset by peer]
15:25  * jdike kills another ubd mmap bug
15:27 < jdike> oooh, looks like that killed two bugs
15:27 < jdike> gotta love collateral damage
15:28 < jdike> heh, make that three bugs
15:28 < green> ;)
15:28 < green> there were 0 bugs in program, programmer fixed one and there became ff bugs, if I remember right ;)
15:29 < jdike> green: hehe
15:29 < jdike> green: and it's worse with 64-bit boxes
15:30 < green> yup
15:30 < green> but you are lucky, because you have only 32 bit
15:30 < green> and in fact you only need to introduce one bug to get rid of a problem ;)
15:30 < jdike> green: heh
15:33 < jdike> Well, ubd mmap looks like it's surviving a kernel build
15:34 < green> cool
15:35 < jdike> oops
15:35 < jdike> partitions/partitions.o: file not recognized: File format not recognized
16:01 < TheDuck> word
16:03 < TheDuck> i need a little help with umld, if anyone has a moment, or some REAL documentation on it  ;)
16:04 < jdike> TheDuck: that would be david, if he's around
16:05 < TheDuck> thanks a lot, btw, great work on UML, this stuff is really impressive
16:05 < TheDuck> i've come from working at a business that was using virtuozzo, what a mess that crap was
16:06 < jdike> TheDuck: tx
16:06 < TheDuck> :-D
16:06 < TheDuck> best of all, its FREE
16:06 < TheDuck> yay opensource!
16:07 < jdike> yeah, virtuozzo is 7K/host or something
16:07 < TheDuck> welp, i know we signed a $1mil contract
16:07 < TheDuck> too much
16:07 < TheDuck> and worst of all, it never worked right, eventually destroying our company
16:09 < TheDuck> lots of disastisfied customers
16:10 < TheDuck> one thing i need to add to umld is some kind of uml migration between hosts, perhaps over sftp or something encrypted of the sort
16:10 < TheDuck> i got a friend working on a REALLY pretty control panel for all of this
16:15 < TheDuck> ps, when we start making some money on the control panel, i'd like to donate something to the core developers of uml
16:19 < jdike> TheDuck: you mean, migrate UMLs from host to host?
16:21 < TheDuck> correct
16:21 < TheDuck> automated by way of web-based interface
16:21 < TheDuck> pick u ml, pick host, and migrate
16:21 < jdike> TheDuck: unless you have some clever plan for doing that, it will require swsusp support in UML
16:22 < jdike> TheDuck: which will happen at some point, it's just not there yet
16:22 < TheDuck> hmm
16:22 < TheDuck> welp, all servers are panned to use the same root_fs
16:22 < TheDuck> just gotta re-create user, and move cow file, no?
16:23 < jdike> TheDuck: oh
16:23 < jdike> TheDuck: I thought you meant moving a running UML
16:23 < TheDuck> hehe, no
16:23 < TheDuck> gotta stop it first
16:23 < TheDuck> moving a running one though,,, damn that would be impressive
16:23 < jdike> TheDuck: yeah, that's pretty easy
16:24 < jdike> TheDuck: it won't be that hard
16:24 < TheDuck> thats like getting your transmission stolen from your car while sitting at a red light
16:24 < TheDuck> lol
16:24 < jdike> TheDuck: suspend it do disk, copy the memory image to new host, restart it
16:24 < TheDuck> thats my thought
16:24 < jdike> TheDuck: the host needs to play some tricks to make sure existing tcp connections move too
16:25 < TheDuck> then delete old stuff, and update database as to where the uml is now located
16:25 < TheDuck> this should only be necessary on those that use lots of disk space, or need more memory, cpu, etc
16:26 -!- jeffro [~Jeff@Ottawa-HSE-ppp244812.sympatico.ca] has joined #uml
16:26 < jdike> TheDuck: swsusp will be good for load balancing between hosts
16:26 < TheDuck> interesting
16:26 -!- jeffro [~Jeff@Ottawa-HSE-ppp244812.sympatico.ca] has left #uml [Leaving]
16:27 < jdike> TheDuck: you might also be able to keep mostly-idle UMLs on disk and only wake them up when there's traffic they need to handle
16:27 < TheDuck> welp, they will all be live webservers
16:28 < TheDuck> dont want to make them sleep
16:28 < TheDuck> customers dont like that
16:50 -!- G2 [~slim_jim@212.159.59.85] has joined #uml
16:51 -!- G2 [~slim_jim@212.159.59.85] has left #uml []
17:57 -!- ghimlet [~hy@62.211.157.30] has joined #uml
17:57 < ghimlet> hy 
18:05 < TheDuck> word  :-D
18:05 < ghimlet> :) 
18:38 -!- kfrench [~kfrench@pcp701781pcs.bowie01.md.comcast.net] has joined #uml
18:48 -!- ghimlet [~hy@62.211.157.30] has quit [Quit: Uscita dal client]
18:51 -!- jdike [~jdike@12.6.236.73] has quit [Quit: Leaving]
19:58 -!- kfrench [~kfrench@pcp701781pcs.bowie01.md.comcast.net] has quit [Read error: Connection reset by peer]
20:15 -!- tarball [~peter@duvel.drunkcoders.com] has joined #uml
20:15 -!- pirlouit [~peter@duvel.drunkcoders.com] has quit [Read error: Connection reset by peer]
20:23 -!- kfrench [~kfrench@pcp701781pcs.bowie01.md.comcast.net] has joined #uml
20:23 -!- coryb [~cory@24.124.36.198] has joined #uml
20:23 -!- kfrench_ [~kfrench@pcp701781pcs.bowie01.md.comcast.net] has joined #uml
20:24 -!- kfrench [~kfrench@pcp701781pcs.bowie01.md.comcast.net] has quit [Client Quit]
20:24 -!- kfrench_ [~kfrench@pcp701781pcs.bowie01.md.comcast.net] has quit [Client Quit]
20:26 -!- tarball [~peter@duvel.drunkcoders.com] has quit [Ping timeout: 501 seconds]
20:28 -!- Ndiin [~eandres@bosporos.mars.asu.edu] has joined #uml
20:29 < Ndiin> Hey.. has anyone ever seen a UML hang *only* after doing a 'reboot'? (hanging right after 'POSIX conformance testing by UNIFIX')
20:29 < Ndiin> The first time through goes fine
20:31 < Ndiin> ah.. hmm
20:31 < Ndiin> POSIX conformance testing by UNIFIX
20:31 < Ndiin> os_set_fd_async : Failed to fcntl F_SETOWN (or F_SETSIG) fd 8 to pid -2145485772, errno = 1
20:31 < Ndiin> All CPUs are go!
20:31 < Ndiin> that happens the first time
20:31 < Ndiin> but it doesn't hang.
21:00 < TheDuck> mine WAS doing that too
21:00 < TheDuck> but it seemed to have corrected itself
21:00 < TheDuck> i wish i knew what i did
21:00  * TheDuck shrugs
21:19 < Ndiin> heh
21:19 < Ndiin> Well, I really need a fix for that.. hmm
22:29 -!- pixelfairy [~pixel@64.81.36.140] has left #uml []
23:07 -!- jamIe [~lara_sk@adsl-68-89-163-233.dsl.hstntx.swbell.net] has joined #uml
23:08 < jamIe> hi, has anyone been able to copy all the files from one mounted ubd to another (formatted with reiserfs?)
23:09 < jamIe> is there anyway to copy (all) the files from one reiserfs to another without screwing the journal on the destination? i'm trying to move my files to a larger filesystem on a boot filesystem.
23:14 -!- jamIe [~lara_sk@adsl-68-89-163-233.dsl.hstntx.swbell.net] has quit [Quit: Client exiting]
23:38 -!- com4 [~com4@12-209-152-183.client.attbi.com] has joined #uml
23:39 -!- pirlouit [~peter@duvel.drunkcoders.com] has joined #uml
23:50 -!- SqueezeCandy [Terry@ACC9B3B2.ipt.aol.com] has joined #uml
23:53 -!- SqueezeCandy [Terry@ACC9B3B2.ipt.aol.com] has quit [Quit:  HydraIRC -> http://www.hydrairc.com <- \o/]
--- Log closed Fri Aug 29 00:00:00 2003