You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@spamassassin.apache.org by bu...@bugzilla.spamassassin.org on 2007/04/12 14:08:07 UTC
[Bug 5413] New: Child processes not being fully killed
http://issues.apache.org/SpamAssassin/show_bug.cgi?id=5413
Summary: Child processes not being fully killed
Product: Spamassassin
Version: 3.1.8
Platform: All
OS/Version: Linux
Status: NEW
Severity: normal
Priority: P5
Component: spamc/spamd
AssignedTo: dev@spamassassin.apache.org
ReportedBy: jhorne@plymouth.ac.uk
We are running 2 servers using Fedora Core 4, and spamassassin 3.1.8.
We have noticed that the child processes are being marked as killed, but that
they actually still exist. As such eventually we find all the child processes in
a killed state, and spamd cannot do any actual work.
We had a similar problem last year, but this was due to SELinux. Both servers
currently have SELinux disabled. See
http://marc.info/?l=spamassassin-users&m=115763870219401&w=2
The problem only seems to occur when the server becomes busy, but even then does
not always happen. The log file shows for example:
Apr 12 11:46:10 mary spamd[16566]: prefork: child states: KIIKKK
I currently have a process monitoring this to warn me when 3 'K's appear, so
neither server should actually stop processing spam. Our max-children is set to
10, and a minimum of 4. Max-conn-per-child is set to 100.
An strace of a 'stuck' child shows:
=========================================================
[root@mary ~]# strace -Ff -p 25131
Process 25131 attached - interrupt to quit
select(16, [10], NULL, NULL, {258, 808000}) = 1 (in [10], left {149, 268000})
read(10, "P....\n", 6) = 6
read(10, 0xb20ba00, 6) = -1 EAGAIN (Resource temporarily
unavailable)
time(NULL) = 1176374820
select(16, [10], NULL, NULL, {300, 0}) = 1 (in [10], left {149, 628000})
read(10, "P....\n", 6) = 6
read(10, 0xb20ba00, 6) = -1 EAGAIN (Resource temporarily
unavailable)
time(NULL) = 1176374971
=========================================================
This just seems to loop round forever. Using 'kill' gets rid of the child.
I have looked through some of the other bug reports, but can find none quite
like this one. The log file shows no error messages such as sysread/syswrite
errors or timeouts.
John.
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
[Bug 5413] Child processes not being fully killed
Posted by bu...@bugzilla.spamassassin.org.
http://issues.apache.org/SpamAssassin/show_bug.cgi?id=5413
------- Additional Comments From rosenbaumlm@ornl.gov 2007-08-28 06:59 -------
I'm seeing the same thing with SpamAssassin v3.2.3, Perl 5.8.8, on Solaris 9:
email# pstack 26988
26988: /usr/local/bin/perl -T /usr/local/bin/spamd -d -u spamd -r /var/run/sp
ff21e5a0 poll (ffbff7f0, 1, 493e0)
ff1d2628 _select (493e0, ff23f310, 0, ff23f310, 14ae718, 0) + 2ec
fe89e1c4 select (10, 14ae718, 0, 0, ffbff968, 800000) + 6c
000c06f8 Perl_pp_sselect (4, ffbff970, 2, ffbff970, cafd70, 4) + 33c
000844e8 Perl_runops_standard (120c00, 0, 0, 0, 123c00, 8d1268) + 30
000286f0 S_run_body (1, ffbff320, 2035c, 12acf8, a64, 1240f0) + 118
00028338 perl_run (123cc0, 24fd8, 10, ffbffc04, 1, 2) + 8c
00024fac main (10, ffbffc04, ffbffc48, 123a88, 0, 0) + a0
00024da4 _start (0, 0, 0, 0, 0, 0) + 5c
email# truss -f -p 26988
26988: poll(0xFFBFF7F0, 1, 300000) (sleeping...)
26988: poll(0xFFBFF7F0, 1, 300000) = 1
26988: read(10, " P . . . .\n", 6) = 6
26988: read(10, 0x014623A0, 6) Err#11 EAGAIN
26988: time() = 1188309227
26988: poll(0xFFBFF7F0, 1, 300000) (sleeping...)
26988: poll(0xFFBFF7F0, 1, 300000) = 1
26988: read(10, " P . . . .\n", 6) = 6
26988: read(10, 0x014623A0, 6) Err#11 EAGAIN
26988: time() = 1188309380
26988: poll(0xFFBFF7F0, 1, 300000) (sleeping...)
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
[Bug 5413] Child processes not being fully killed
Posted by bu...@bugzilla.spamassassin.org.
http://issues.apache.org/SpamAssassin/show_bug.cgi?id=5413
------- Additional Comments From jhorne@plymouth.ac.uk 2007-09-11 03:13 -------
We have just upgraded both servers to CentOS 5 and SpamAssassin 3.2.3, perl 5.8.8.
The problem has not occurred since then (over a week now). Previously the
problem occurred every day or so (I ran a monitoring program which reported if 3
or more 'K's appeared in the mail log file). The servers have come under a quite
heavy load, but no sign of any half-killed processes.
Not sure if you want to close this call, as it seems fixed for me, or leave it
open if there is still an issue under Solaris 9. Obviously I cannot say whether
the 'fix' was because of SpamAssassin, perl or perhaps something in CentOS?
John.
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.