Let me update you that we are past the issue in a number of ways.
First, we upgraded to 0.9.9 Sunday night, found that we could run 36+
hours, but then ran out of swap (32MB main memory + 8GB swap was
configured). The swap is now sized at 46GB. You may, under extreme
and prolonged load, have an issue but we are unlikely to be able to
recreate this scenario.
The root cause of our issue was a very large amount of I/O being
driven through the system due to an apparent malfunction in our own
software higher up the stack. We are working around that issue and
consider tgt to be working fine with this most recent release.
Bottom line - the segfault was resolved by the 0.9.9 release.
Thanks!
Marty
FUJITA Tomonori wrote:
Very sorry for the late response,
On Fri, 25 Sep 2009 07:17:31 -0700 (PDT)
Martin Montreuil <martin.montreuil@xxxxxxxxxx> wrote:
Under V0.9.7 received:
Sep 24 05:12:57 storageserver kernel: tgtd[31665]: segfault at 0000555e4ee57d90 rip 0000003dc14715a8 rsp 00007fff7f899ce0 error 4
Upgraded to V0.9.8 and:
Sep 25 01:37:02 storageserver kernel: tgtd[31609]: segfault at fffffffffffffff0 rip 0000000000405ae4 rsp 00007fffc7fe3940 error 4
Not repeatable but happens generally within 24 hours. There are 143 disks being served in this configuration and multipathd is in use. This is a new installation.
targets.conf looks like:
<target iqn.2009-06.crrel:storageserver.disks>
backing-store /dev/mapper/mpath1
backing-store /dev/mapper/mpath2
...
backing-store /dev/mapper/mpath143
allow-in-use yes
</target>
Under V0.9.8 we are also seeing something new - BUG: soft lockup's (below) just prior to the segfault.
Any suggestions?
Firstly, can you try the latest git tree?
You hit a kernel bug (soft lockup) so I guess that an initiator drops
a connection then tgtd crashes.
Seems that there might be a bug about cleaning up unfinished
tasks. I'll dig into the code later. I have several things to do
before the kernel summit so it might take some time. Sorry about that.
--
To unsubscribe from this list: send the line "unsubscribe stgt" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe stgt" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
