Re: Cron <buildd@kullervo> update-buildd.net status
On Mon, Feb 11, 2013 at 03:19:14PM +0100, Andreas Schwab wrote:
> "Christian T. Steigies" <cts@debian.org> writes:
>
> > On kullervo it happens about twice a day, mostly in some buildd scripts (and
> > not during the actual build, I hope). Can you suggest something to look at
> > this in detail?
>
> As a first step you should log as much as possible of the context when
> the sigill happens.
There is not much to capture, I just receive several emails with failed
commands:
Subject: Cron <buildd@kullervo> /usr/bin/buildd-watcher
Illegal instruction
Subject: Cron <buildd@kullervo> /usr/bin/buildd-uploader
Bus error
Subject: Cron <buildd@kullervo> update-buildd.net status
/usr/local/bin/update-buildd.net: line 337: 31452 Illegal instruction pgrep -u $BUILDDUSER -f 'sbuild.*--dist=' &>/dev/null
Subject: Cron <buildd@kullervo> update-buildd.net status
dirname: missing operand
Try 'dirname --help' for more information.
Subject: Cron <buildd@kullervo> /usr/bin/buildd-watcher
Segment violation
Subject: Cron <buildd@kullervo> update-buildd.net status
Signal 11 (SEGV) caught by ps (procps-ng version 3.3.4).
ps:display.c:59: please report this bug
Subject: Cron <buildd@kullervo> update-buildd.net status
Signal 4 (ILL) caught by ps (procps-ng version 3.3.4).
ps:display.c:59: please report this bug
However, the frequency of these messages increased, instead of 2 per day
yesterday there were more than 10. In the end, I captured this from the
console when the machine froze:
BUG: soft lockup - CPU#0 stuck for 22s! [buildd-uploaded:9888]
Modues linked in: dm_snapshot ipv6 dm_mod loop rtc_rp5c01 zorro8390
parport_amiga evdev parport dmasound_paula dmasound_core soundcore
Format 00 Vector: 0078 PC: 0000608e Status: 2000 Not tainted
ORIG_DO: ffffffff D=: 062f6000 A2: 061e6a00 A1: c02ece98
A0: 204b200b D5: 00000018 D4: 80005584
D3: c016c382 D2: 00000006 D1: 00000003
The load was pretty high, also the memory usage was high with a lot of swap
used, so perhaps the memory is bad or there is a problem with the swap?
All in all maybe not quite ready for MNNM yet.
Christian
Reply to: