Back to Home / #uml / 2007 / 02 / Prev Day | Next Day
#uml IRC Logs for 2007-02-20

---Logopened Tue Feb 20 00:00:43 2007
00:04|-|tasaro [] has quit [Server closed connection]
00:04|-|tasaro [] has joined #uml
00:06|-|linbot [] has quit [Server closed connection]
00:06|-|linbot` [] has joined #uml
00:09|-|caker [] has quit [Server closed connection]
00:09|-|caker [] has joined #uml
01:37|-|mike_m [~mike@] has quit [Quit: Client exiting]
01:38|-|motp [~motp@] has joined #uml
02:06|-|motp [~motp@] has quit [Quit: Leaving]
02:08|-|Ancalagon [] has joined #uml
03:01|-|kokoko1 [~Slacker@] has quit [Ping timeout: 480 seconds]
03:41|-|tyler [~tyler@] has joined #uml
03:45|-|richardw [] has joined #uml
03:59|-|arun [~arun@chobie.cs.Virginia.EDU] has quit [Ping timeout: 480 seconds]
04:06|-|tasaro_ [] has joined #uml
04:06|-|tasaro [] has quit [Write error: connection closed]
04:24|-|arun [~arun@chobie.cs.Virginia.EDU] has joined #uml
05:08|-|DeHackEd [] has quit [Ping timeout: 480 seconds]
06:11|-|pgstudy [] has joined #uml
06:19|-|writerz_ [] has joined #uml
06:26|-|mode2_ [] has quit [Ping timeout: 480 seconds]
06:37|-|albertito [] has quit [Read error: Connection reset by peer]
06:42|-|albertito [] has joined #uml
06:49|-|Diablo-D3 [] has joined #uml
06:49<Diablo-D3>A HOWTO on building a Debian UML image for Debian
07:35|-|tasaro_ changed nick to tasaro
07:40|-|richardw_ [] has joined #uml
07:44|-|richardw [] has quit [Read error: Operation timed out]
08:00|-|baroni [] has joined #uml
08:18|-|flatronf700B [~flatronf7@] has quit [Ping timeout: 480 seconds]
08:31|-|DeHackEd [] has joined #uml
08:40|-|richardw_ [] has quit [Quit: Leaving]
08:41|-|Ancalagon [] has quit [Quit: Chatzilla 0.9.75 [Firefox 1.0.4/20070116]]
09:14|-|Diablo-D3 [] has quit [Ping timeout: 480 seconds]
09:17|-|albertito [] has quit [Quit: chau!]
09:46|-|silug_ [~steve@] has quit [Ping timeout: 480 seconds]
09:48|-|jdike [] has joined #uml
09:48<jdike>Hi guys
09:52<dgraves>morning jdike.
09:52<dgraves>how's kicks/
09:52<dgraves>you have a good holiday?
09:52<jdike>you mean those random days when the mail doesn't show up?
09:53<DeHackEd>good morning. is it a holiday south of the border?
09:53<jdike>yesterday, supposedly
09:55<DeHackEd>anyways, your patch booted up but those extra UML processes (IO related I guess) basically raped my CPU. I'm going to try tampering with the VSDO setting next.
09:55<dgraves>jdike: yeah. that was my thought too. :)
09:56[~]dgraves doesn't want to picutre DeHackEd's cpu now...
09:56<dgraves>that sounds painful.
09:57<DeHackEd>dgraves: it is. 100% CPU usage by one of those IO threads during what would be busy IO work by the UML. of course the "kernel mode" process was kept busy as well. in all UML often used 170% CPU
09:57<jdike>DeHackEd, what is the UML doing?
10:00<DeHackEd>logging in (bash doing its thing I guess), downloading and unzipping files.
10:03<dgraves>DeHackEd: try killing your system idle thread. :)
10:03<DeHackEd>been there, done that.
10:04<jdike>what does ps show inside UML?
10:04<DeHackEd>exactly what i'd have expected
10:10<jdike>nothing unusual running then?
10:10<jdike>what does it look like on the host?
10:10<DeHackEd>I just rebooted with a kernel with VSDO compat mode disabled.
10:11<DeHackEd>when it went nuts, the "last" UML process created would use 100% CPU, almost all system time. The "first" (kernelspace) process would use around 60% CPU. If it lasted a long time, the soft lockup warning would fire...
10:12|-|kybe___ [] has joined #uml
10:13<jdike>can you strace the 100% system time one?
10:14<DeHackEd>there was a lot of poll()ing, read/write of one byte...
10:15<jdike>there's a fix for that
10:24|-|kokoko1 [~Slacker@] has joined #uml
10:27<DeHackEd>Nice. That seems to have done the trick.
10:28<DeHackEd>the kernel VSDO option in the processor-type config for the host kernel fixes the previous problem from yesterday.
10:30<DeHackEd>(disabling the compatability mode is the fix)
10:32<jdike>and my patch also fixes the problem, correct?
10:32<DeHackEd>it did
10:33<DeHackEd>it hung eventually, but that was with the IO problem as well. the kernel thread got stuck on a read() operation...
10:33<DeHackEd>so I was going around trying to find alternatives.
10:44|-|fbenites [~chatzilla@] has joined #uml
11:04|-|AllenJB_ [] has joined #uml
11:05|-|AllenJB_ [] has quit []
11:05[~]jdike sends a few patches off to Andrew
11:05<jdike>including the VDSO one
11:06<jdike>and one minute after sending that, I remember two other places I should have sent it to
11:09<kokoko1>jdike, hi
11:18<kokoko1>IMMEDIATE action, remove all access of YYY
11:18<kokoko1>I hate to see these emails :(
11:18<kokoko1>which mean firing :(
11:24|-|ram [] has quit [Ping timeout: 480 seconds]
11:38|-|silug_ [~steve@] has joined #uml
11:46|-|tyler [~tyler@] has quit [Ping timeout: 480 seconds]
11:53|-|tyler [~tyler@] has joined #uml
11:59<dgraves>jdike: yay! i'm gonna get more spam! :)
12:02<jdike>the rewards of being a kernel contributor
12:09<dgraves>its odd that someone would actually harvest it.
12:09<dgraves>you'd think they'd at least target the spam... :-P
12:10<jdike>I'm getting spam with $SUBJECT == line from random Linux HOWTO
12:10<jdike>so there is some targetting happening
12:10<dgraves>yeah, but what about like geek toys?
12:11<dgraves>or geek women nude?
12:11<dgraves>or something that might actually mean something to us. :)
12:11<dgraves>or cisco 7970 phones!
12:11<dgraves>(we just got them.
12:11<dgraves>they have a 3 inch lcd on them.
12:11<dgraves>i sat down at my phone and tried to work this morning. :-P
12:12<jdike>stick it under a microscope
12:12<dgraves>anyway, i'm bouncin.
12:22|-|ram [] has joined #uml
12:25|-|fbenites [~chatzilla@] has quit [Remote host closed the connection]
12:37|-|jeroenh [] has joined #uml
12:42[~]jdike missed sending those patches to uml-devel
12:42<jdike>mistyped the address
12:55|-|kos_tom [] has joined #uml
13:28<jeroenh>jdike: about the uml_switch silent mode: it is actually generating a lot of output when there is a loop in the network
13:28<jeroenh>I think that this might even be a bug
13:29<jdike>what's the output?
13:29<jeroenh>a lot of: "send_sock sending to fd 4 Resource temporarily unavailable"
13:29<jeroenh>and it's moving ports about
13:29<jeroenh> Addr: fe:fd:00:00:00:02 New port 5 old port 7
13:30<jdike>is it seeing MACs move from port to port?
13:30<jdike>i.e. is traffic from a given MAC appearing on different ports?
13:31<jeroenh>that would be the case with a loop in the network, yes
13:33<jdike>then having it update its idea of where to send such traffic would seem to be correct
13:34<jeroenh>yes, that's true
13:34<jeroenh>but what about the fd 4 messages?
13:35<jdike>those are just telling you it's dropping packets because a kernel queue is full
13:35<jdike>I think that can be improved
13:35<jdike>I think it's using one descriptor to send out all packets
13:36<jeroenh>ah yes
13:37<jdike>you could have one per port
13:37<jdike>that should cut down on those messages
13:37<jdike>and not have one slow UML cause dropped packets for all the others
13:37<jeroenh>by the w
13:38<jeroenh>by the way, it really does have effect on the terminal that started it
13:38<jeroenh>I'm issuing commands over ssh, and if something like this occurs, I'm missing keystrokes
13:39<jdike>when output is redirected to /dev/null?
13:39<jeroenh>and it's not the machine being under heavy load, other terminals work fine
13:40<jdike>there's nothing unusual about the switch
13:40<jdike>it prints stuff to stdout just like anything else
13:41<jdike>unless it's sending packets to fd 0 or something strange
13:42<jeroenh>I don't know...I just know that when I am running a uml_switch from that terminal I'm getting that behavior
13:42<jdike>I don't see any other way it could confuse the terminal
13:43<jeroenh>and if I run it using a separate daemonize script, then I don't get that behavior
13:45|-|hfb [] has joined #uml
13:53|-|hfb [] has left #uml [Leaving]
14:00<jeroenh>oh and I have some more bugreports for you jdike
14:01<jdike>it's such an innocent piece of code :-)
14:01<jeroenh>if you use "home=x" as parameter for uml it will crash most of the time
14:01<jeroenh>at least, does that
14:02<jeroenh>and doing host mount on a non-existant host directory will also crash uml
14:02<jdike>boots here
14:02<jdike>that last has been fixed
14:03<jeroenh>in 2.6.20 ?
14:03<jdike>home isn't a UML option
14:03<jeroenh>I know
14:03<jdike>it has no effect whatsoever here
14:03<jeroenh>hmm, now that I think about it, this may have been caused by the non-existing host directory
14:04<jeroenh>(yes, I had a weird startup script)
14:04<jeroenh>yeah, it's probably that
14:05<jeroenh>but the silent uml_switch would make it a lot easier to use uml in scripts
14:08<jdike>it'll be easy enough to go through it and rip out the printfs
14:08<jdike>and not too much harder to put them under the control of a -silent switch
14:08<jdike>or -q
14:09<jeroenh>cool, that almost sounds like something that even I could do
14:11[~]kokoko1 is gone
14:13<jdike>especially the "Resource temporarily unavailable" stuff
14:13<jdike>there isn't a lot of value to that
14:19<jeroenh>jdike: the latest version is the one from April 2004?
14:21<jdike>no, look at
14:28|-|richardw [] has joined #uml
14:42<jeroenh>I must confess, I don't really know C and I wouldn't really know how to pass the silent option between uml_switch.c to port.c
14:42<jeroenh>I guess you'd have to change the call to new_port to include an extra argument
14:45<jdike>rather than pass a quiet switch everywhere, you could also just make it a global
15:20|-|richardw [] has quit [Quit: Leaving]
15:26|-|mode2_ [] has joined #uml
15:32|-|writerz_ [] has quit [Ping timeout: 480 seconds]
15:35[~]jeroenh pasted
15:35<jeroenh>that's a start
15:37<jeroenh>however, it still messes up my terminal: if I launch uml_switch quietly and in daemon mode, then create a uml connected to that, connect to it using a screen, and bring up the interface, the screen session freezes
16:12|-|tarsin [] has quit [Server closed connection]
16:12|-|tarsin [] has joined #uml
16:13|-|writerz_ [] has joined #uml
16:18|-|mode2_ [] has quit [Ping timeout: 480 seconds]
16:19<jdike>have you looked at it?
16:19<jdike>there's nothing there
16:22<mjf>jdike: The patch works as charm. No freeze since patching the 2.6.20 w/ it. :-)
16:22<mjf>jdike: (I was 'uroboros', now 'mjf', right?)
16:22<jdike>it's headed for mainline
16:22<mjf>jdike: I wanted to ask about the second patch you suggested to caker. What was this one for?
16:23<mjf>jdike: I wanna to experiment with device-mapper on one of the UMLs, do I need to patch it? I was something concerning DM, if I remember the discussion properly...
16:24<jdike>dm should be fine
16:25<jdike>the second patch being the one yesterday?
16:25<mjf>jdike: Eh? I was not paying attention to this channel yesterday. The last time I was here (the day we were debugging my freezing UML).
16:25<mjf>jdike: (I hope)
16:25<jdike>then remind me what I was talking about :-)
16:26<mjf>Two patches were discusses. The one that was initiated by the freezing of my UMLs and the caker's problem. The former was solved with just a very tiny patch (few lines only).
16:28<mjf>I thing it was about some kernel panic while using DM or something similar. But I was not paying attention so much to get the all information (because I was trying to learn something about GDB during the evening and thus not reading the this channel all the time while reading some howtos...)
16:29<mjf>Hm, and then I accidentaly breaked my IRSSI and lost the whole log. :-)
16:30[~]jeroenh pasted
16:30<jeroenh>there you go
16:30<jeroenh>sorry about the first paste, used the wrong command from textmate
16:37<jdike>Oh, the LVM patch
16:38|-|tyler [~tyler@] has quit [Ping timeout: 480 seconds]
16:38<mjf>jdike: perhaps, yes... do you still have some URL to it? I wanna to check what it is for... :)
16:38[~]mjf really needs some good bookmarking system ;)
16:41<jdike>jeroenh, that looks workable
16:42<jdike>it might be cleaner to have a output() function which behaves like printf, but checks the value of silent
16:43|-|linbot` changed nick to linbot
16:43<jeroenh>but this is my first serious stab at coding in C++
16:43<jeroenh>I've mainly coded in Python so far
16:44<mjf>I am just starting to learn ANSI C ;)
16:44<jdike>jeroenh, it's C
16:44<mjf>Mostly coding in BASH and Perl before, but falled in love with C last few weeks... :)
16:45<jeroenh>see, shows what I know
16:46<jeroenh>but I was mainly wondering what would disrupt my terminal session now that the print statement is gone
16:46<jeroenh>and/or eating my keystrokes
16:47<jeroenh>but it's really bed time for me now, talk to you later
16:49<mjf>jdike: thanks
16:50<jeroenh>hah, bed time reading huh ?
16:50<mjf>jdike: should I use it?
16:50<mjf>jeroenh: yes, of course :)
16:50|-|jeroenh [] has quit [Quit: jeroenh]
16:51<mjf>cya == ciao?
16:51<jdike>see ya
16:51<jdike>more literally
16:51<mjf>aha, "C" ya
16:51<jdike>is UML crashing when you use DM?
16:51<mjf>not at the very moment. But I wanna to to do really *bad* things wit it...
16:52<mjf>perhaps there will be a raid too in play ;)
16:52<mjf>a lot of levels...
16:53<mjf>the "nbd" stands for "net-block device"
16:53<mjf>LVM DM-crypted network RAID ;)
16:53<mjf>to be precise.
16:54<mjf>I am just experimenting. UML seems to be the best way how to not corrupt some running systems... :-)
16:55<mjf>jdike: Will this patch go to the next release as well?
16:55<mjf>OK. I will keep experimenting without it until new release will be released...
16:56<mjf>You know, I really love the way you can put one layer over another with Linux. That's simply GREAT!
16:59<mjf>I go sleeping. Good night.
17:02|-|nessie [] has quit [Ping timeout: 480 seconds]
17:08|-|kos_tom [] has quit [Ping timeout: 480 seconds]
17:33<fo0bar>woo, just ordered the hardware and got service offerings worked out for client UML hosting
17:33<fo0bar>caker: don't worry, we're totally different market segment :)
17:35<fo0bar>clients don't come to us for hosting/colo/etc, they come for a package of marketing campaigns/comp design/website design+hosting/advertising/etc
17:36<fo0bar>UML is being marketed as a tier between shared hosting (die die die) and full-on dedicated hosting
17:37<fo0bar>3.6ghz PD/4G/500GB usable in RAID10, split 8 ways
17:59|-|Diablo-D3 [] has joined #uml
18:31|-|jdike [] has quit [Quit: Leaving]
19:00|-|Hunger [] has quit [Server closed connection]
19:08|-|silug_ [~steve@] has quit [Read error: Connection reset by peer]
19:13|-|flatronf700B [~flatronf7@] has joined #uml
19:20|-|Hunger [] has joined #uml
19:26|-|richardw [] has joined #uml
19:27|-|ram [] has quit [Ping timeout: 480 seconds]
19:55|-|richardw [] has quit [Quit: Leaving]
20:32|-|arun [~arun@chobie.cs.Virginia.EDU] has quit [Server closed connection]
20:32|-|arun [~arun@chobie.cs.Virginia.EDU] has joined #uml
21:04|-|silug_ [~steve@] has joined #uml
21:14|-|ousado_ [] has joined #uml
21:16|-|ram [] has joined #uml
21:22|-|ousado__ [] has quit [Ping timeout: 480 seconds]
21:53|-|silug_ [~steve@] has quit [Read error: Connection reset by peer]
21:58|-|nessie [] has joined #uml
22:41|-|Nem^1 [] has joined #uml
22:44|-|Nem^ [] has quit [Read error: Operation timed out]
22:45|-|Nem^1 changed nick to Nem^
22:57|-|Netsplit <-> quits: flatronf700B, writerz_, Nem^, ram
23:00|-|baroni [] has quit [Quit: :wq]
23:00|-|VS_ChanLog [] has left #uml [Rotating Logs]
23:00|-|VS_ChanLog [] has joined #uml
23:00|-|Netsplit over, joins: flatronf700B, Nem^, ram, writerz_
23:49|-|silug_ [~steve@] has joined #uml
---Logclosed Wed Feb 21 00:00:41 2007