Date: Tue, 15 Mar 2005 08:22:51 -0500 From: Michael Boers <msb@datacompusa.com> To: Pawel Jakub Dawidek <pjd@FreeBSD.org> Cc: freebsd-geom@FreeBSD.org Subject: Re: gmirrored boot drives locks up during buildworld Message-ID: <42cc591ca3faf6a5ab49fa7101e5945c@datacompusa.com> In-Reply-To: <20050314190650.GX9291@darkness.comp.waw.pl> References: <f7944513206a01f926630fed59abc188@datacompusa.com> <20050314190650.GX9291@darkness.comp.waw.pl>
next in thread | previous in thread | raw e-mail | index | archive | help
Pawel,
Thank-you for your quick response.
In addition to enabling HTT in the bios, I have made the following
changes to the kernel,
options SMP
options MPTABLE_FORCE_HTT
options MP_WATCHDOG
and set the following flags
kern.geom.mirror.debug=1
debug.watchdog=1
I reran the "while (true) do make clean; make buildworld; done" test
and the machine has locked up as before.
I was monitoring the test with a terminal running top
last pid: 57530; load averages: 3.79, 3.24, 2.93
up 0+07:18:14 23:31:10
148 processes: 6 running, 111 sleeping, 31 waiting
CPU states: 31.3% user, 0.0% nice, 18.5% system, 0.2% interrupt,
50.0% idle
Mem: 110M Active, 385M Inact, 89M Wired, 672K Cache, 112M Buf, 416M Free
Swap: 512M Total, 512M Free
PID USERNAME PRI NICE SIZE RES STATE C TIME WCPU CPU
COMMAND
11 root 130 0 0K 12K CPU1 1 383:33 100.54% 100.54%
mp_watchdog cpu 1
3 root -8 0 0K 12K - 0 21:40 29.74% 29.74% g_up
29 root -64 -183 0K 12K WAIT 0 2:42 5.22% 5.22%
irq18: uhci2+
4 root -8 0 0K 12K - 0 3:27 2.05% 2.05%
g_down
12 root 96 0 0K 12K RUN 0 383:56 0.00% 0.00%
idle: cpu0
576 mysql 20 0 56896K 34584K kserel 0 1:17 0.00% 0.00%
mysqld
713 root 8 0 2000K 1456K nanslp 0 0:19 0.00% 0.00%
gstat
25 root -64 -183 0K 12K WAIT 0 0:19 0.00% 0.00%
irq14: ata0
50 root -8 0 0K 12K m:w1 0 0:18 0.00% 0.00%
g_mirror boot
37 root -28 -147 0K 12K WAIT 0 0:15 0.00% 0.00%
swi5: clock sio
53 root 171 52 0K 12K RUN 0 0:14 0.00% 0.00%
pagezero
680 root 96 0 2440K 1728K CPU0 0 0:13 0.00% 0.00% top
579 mysql 20 0 132M 43196K kserel 0 0:09 0.00% 0.00%
mysqld
577 mysql 20 0 56576K 24688K kserel 0 0:08 0.00% 0.00%
mysqld
39 root 76 0 0K 12K - 0 0:06 0.00% 0.00%
yarrow
594 msb 96 0 6216K 3020K select 0 0:05 0.00% 0.00% sshd
56 root 20 0 0K 12K syncer 0 0:03 0.00% 0.00%
syncer
and a terminal running gstat
dT: 0.510 flag_I 500000us sizeof 240 i -1
L(q) ops/s r/s kBps ms/r w/s kBps ms/w %busy Name
0 0 0 0 0.0 0 0 0.0 0.0| ad0
0 0 0 0 0.0 0 0 0.0 0.0|
mirror/boot
0 0 0 0 0.0 0 0 0.0 0.0| ad0a
0 0 0 0 0.0 0 0 0.0 0.0| ad0b
0 0 0 0 0.0 0 0 0.0 0.0| ad0c
0 0 0 0 0.0 0 0 0.0 0.0| ad0d
0 0 0 0 0.0 0 0 0.0 0.0| ad0e
0 0 0 0 0.0 0 0 0.0 0.0| ad0f
0 0 0 0 0.0 0 0 0.0 0.0|
mirror/boota
0 0 0 0 0.0 0 0 0.0 0.0|
mirror/bootb
0 0 0 0 0.0 0 0 0.0 0.0|
mirror/bootc
0 0 0 0 0.0 0 0 0.0 0.0|
mirror/bootd
0 0 0 0 0.0 0 0 0.0 0.0|
mirror/boote
0 0 0 0 0.0 0 0 0.0 0.0|
mirror/bootf
0 0 0 0 0.0 0 0 0.0 0.0| ad1
0 0 0 0 0.0 0 0 0.0 0.0| acd0
0 0 0 0 0.0 0 0 0.0 0.0| ad4
0 0 0 0 0.0 0 0 0.0 0.0| ad6
0 0 0 0 0.0 0 0 0.0 0.0| ad4c
0 0 0 0 0.0 0 0 0.0 0.0| ad4d
0 0 0 0 0.0 0 0 0.0 0.0| ad6c
0 0 0 0 0.0 0 0 0.0 0.0| ad6d
I have not touched the machine yet. Is there any other info I can
provide?
--
Michael Boers
Datacomp
On Mar 14, 2005, at 2:06 PM, Pawel Jakub Dawidek wrote:
> On Mon, Mar 14, 2005 at 01:46:15PM -0500, Michael Boers wrote:
> +> I recently installed FreeBSD 5.3 on a machine to be my primary mysql
> +> server. The machine failed after about 3 weeks of heavy use. The
> +> machine did not panic, it just froze and some random characters
> +> appeared on the console. A reboot restored the system for another
> few
> +> weeks. On the third failure I took it out of production.
> +>
> +> The machine consists of a Intel Pentium 4 EE HT with a pair of 80
> +> gigabyte IDE gmirrored boot drives and a pair of 250 gigabyte IDE
> +> gmirror data drives.
> +>
> +> With the machine out of production, I used
> +>
> +> while (true) do make clean; make buildworld; done
> +>
> +> to exercise the machine until it failed. Usually within three
> days. I
> +> swapped video cards, memory, hard drives, and played with bios
> settings
> +> to no avail. Finally I determined that when I ran without using
> +> gmirror, the machine would build indefinitely.
> +>
> +> Finally, I tried the buildworld test on a completely different (amd
> vs
> +> intel, scsi vs ide disks) machine and it failed in less than 3
> hours.
> +>
> +> Because the system freezes rather than panics, I have no diagnostic
> +> information to provide.
> +>
> +> If this is a possible gmirror bug, please let me know if there is
> any
> +> other information I can provide. I am very interested in using
> gmirror
> +> but I want to make sure it is safe. Please feel free to call me at
> the
> +> below number if necessary.
>
> Could increase kern.geom.mirror.debug to 1?
> Could you turn on HTT and compile your kernel with MP_WATCHDOG (you
> should
> also set debug.watchdog to 1)?
>
> --
> Pawel Jakub Dawidek http://www.wheel.pl
> pjd@FreeBSD.org http://www.FreeBSD.org
> FreeBSD committer Am I Evil? Yes, I Am!
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?42cc591ca3faf6a5ab49fa7101e5945c>