Re: Relating to Domains..

From: Murray S. Kucherawy <msk_at_blackops.org>
Date: Wed, 28 Sep 2011 11:50:20 -0700 (PDT)

On Wed, 28 Sep 2011, N. wrote:
> I'm having a very weird problem trying to get that port 8891 "free". I
> keep trying to kill processes, and more spawn. And its from the opendkim
> user.
>
> Actually, I didn't have this error before and opendkim was starting
> fine. But it just wasn't logging or signing email.
>
> Also, with this instruction Murray:
>
> "2) Start strace watching the child process. Use "ps laxww" to figure
> out which one was the child."
>
> Could you link me to the text that discusses the "child"? I assume
> this is opendkim "child" yes? I used ps laxww but I saw a lot of
> things I wasn't familiar with. How do I identify the "child" in that
> output? (what text will I see)? Also, do you happen to have a command
> for strace handy that I should use to have it running in the
> background? I really appreciate it.

medusa[1127]% ps laxww | fgrep opendkim
  1000 879 1 386 8 0 17608 0 wait IWs ?? 0:00.00 /home/msk/auth/opendkim -x /home/msk/auth/opendkim.conf (opendkim-2.4.2)
  1000 880 879 0 20 0 33944 16776 kserel S ?? 4:31.29 /home/msk/auth/opendkim -x /home/msk/auth/opendkim.conf (opendkim-2.4.2)

The first column is the UID of the process. The second is the pid of the
process, and the third is the pid of its parent. You can see here that
the second one is the child, as its parent is 879 which is also an
opendkim process. The first one's parent is the init process, so it is
not the opendkim child.

For strace, try:

         strace -o strace.out -f -p 880

(replace "880" with the process ID of your opendkim child process)

...and then send a message through postfix.

You are probably seeing 8891 not free because opendkim is configured to
restart the child when it dies. Instead, you need to kill the parent.
Using "kill -9" is too severe to get it to shutdown cleanly unless you use
it on both the parent and the child at the same time. Instead, using
"kill -15" on the parent should complete the process.

On the other hand, you can simplify your work by using "strace" on the
process that "lsof" tells you is holding port 8891, and then run your
postfix test.
Received on Wed Sep 28 2011 - 18:50:38 PST