RE: Opendkim number of threads growing abnormaly under Solaris 10

From: Murray S. Kucherawy <msk_at_cloudmark.com>
Date: Tue, 12 Oct 2010 11:15:17 -0700

> -----Original Message-----
> From: opendkim-dev-bounce_at_lists.opendkim.org [mailto:opendkim-dev-
> bounce_at_lists.opendkim.org] On Behalf Of Christian Pélissier
> Sent: Tuesday, October 12, 2010 2:24 AM
> To: opendkim-dev_at_lists.opendkim.org
> Subject: Opendkim number of threads growing abnormaly under Solaris 10
>
> Under Solaris 10 x86 opendkim number of threads are growing abnormaly.
>
> When launched opendkim and for example another milter such
> clamav-milter have eaxctly the same number of threads.
>
> After running 12 hours opendkim number of threads is higth and clamav
> (and 2 others milter) stay low varying according to the same number of
> sendmail instances.
>
> opendkim = 75 threads
> clamav-milter = 15 threads
>
> pstack command show that many threads stay blocked inside mlfi_eom
> milter callback (I have the same problem with Autorestart option or
> not).
>
> # pstack 29694 | grep mlfi
> 0805c221 mlfi_eoh (8101470, fc314fb0, fef139e4, fc314f70, 0, 8101470)
> +
> 115a
> 0805cf4b mlfi_eom (8149210) + ae1
> 0805cf4b mlfi_eom (80b1640) + ae1
> 0805cf4b mlfi_eom (812a278) + ae1
> 0805cf4b mlfi_eom (80b4f58) + ae1
> 0805cf4b mlfi_eom (80b03f8) + ae1
> 0805cf4b mlfi_eom (8119170) + ae1
> 0805cf4b mlfi_eom (80ae458) + ae1
> 0805cf4b mlfi_eom (80efb00) + ae1
> 0805cf4b mlfi_eom (810aff8) + ae1
> 0805cf4b mlfi_eom (80b5208) + ae1
> 0805cf4b mlfi_eom (80c5938) + ae1
> 0805cf4b mlfi_eom (80b5808) + ae1
> 0805cf4b mlfi_eom (80b5930) + ae1
> 0805cf4b mlfi_eom (80b47f0) + ae1
> 0805cf4b mlfi_eom (80bf720) + ae1
> 0805cf4b mlfi_eom (80b80d0) + ae1
> 0805cf4b mlfi_eom (80d0648) + ae1
> 0805cf4b mlfi_eom (80dfaa0) + ae1
> 0805cf4b mlfi_eom (80b5470) + ae1
> 0805cf4b mlfi_eom (80e5360) + ae1
> 0805cf4b mlfi_eom (80cab28) + ae1
> 0805cf4b mlfi_eom (8116da8) + ae1
> 0805cf4b mlfi_eom (80af830) + ae1
> 0805cf4b mlfi_eom (80ccd68) + ae1
> 0805cf4b mlfi_eom (80af8d0) + ae1
> 0805cf4b mlfi_eom (8109378) + ae1
> 0805cf4b mlfi_eom (80bf638) + ae1
> 0805cf4b mlfi_eom (81307e8) + ae1
> 0805cf4b mlfi_eom (80fb238) + ae1
> 0805cf4b mlfi_eom (80fcf50) + ae1
> 0805cf4b mlfi_eom (8108760) + ae1
> 0805cf4b mlfi_eom (811f278) + ae1
> 0805cf4b mlfi_eom (80fcc90) + ae1
> 0805cf4b mlfi_eom (811d168) + ae1
> 0805cf4b mlfi_eom (8112230) + ae1
> 0805cf4b mlfi_eom (8118d10) + ae1
> 0805cf4b mlfi_eom (8138f70) + ae1
> 0805cf4b mlfi_eom (8162530) + ae1
> 0805cf4b mlfi_eom (814fbc0) + ae1
> 0805cf4b mlfi_eom (8154b60) + ae1
> 0805cf4b mlfi_eom (815b268) + ae1
> 0805cf4b mlfi_eom (8131b48) + ae1
> 0805cf4b mlfi_eom (8162648) + ae1
> 0805cf4b mlfi_eom (8150de8) + ae1
> 0805cf4b mlfi_eom (8139f28) + ae1
> 0805cf4b mlfi_eom (8136c70) + ae1
> 0805cf4b mlfi_eom (8148088) + ae1
> 0805cf4b mlfi_eom (814f2b8) + ae1
> 0805cf4b mlfi_eom (8119040) + ae1
>
> Everytime the following message is logged to syslog :
>
> ... sendmail[10290]: [ID 801593 mail.error] o9C98SZB010290: Milter
> (opendkim): timeout before data read, where=eom
>
> opendkim/threads = opendkim/threads + 1 (relative to clamav-milter)
>
> It looks like a bug.
>
> Other used milter (clamav-milter, milter-greylist, j-chkmail) do not
> have this problem.

We need to see a stack trace in order to understand where it's getting stuck.

Can you attack a debugger like gdb to it and ask for "thread apply all bt"? dbx may have an equivalent function. Unfortunately I have no idea what "ae1" is.
Received on Tue Oct 12 2010 - 18:15:26 PST