Fwd: Multiple innfeed processes

Julien ÉLIE julien at trigofacile.com
Mon Mar 18 19:12:17 UTC 2013


Hi Jesse,

I do not see why innfeed does not properly shut down.  It definitively should stop when innd closes its pipe to innfeed upon reloading.
Maybe people in the inn-workers mailing-list will have ideas for that.

-- 
Julien

-------- Message original --------
Sujet: 	Multiple innfeed processes
Date : 	Fri, 08 Mar 2013 12:10:43 -0600
De : 	Jesse Rehmer
Pour : 	Julien ÉLIE

Julien,

I sorted out all of the weird CNFS issues I was having previously.  Turned out to be an issue on the SAN which caused a lot of strange server level issues on the reader machine.  I do have an issue on my feeder machine that I'm having trouble diagnosing/resolving.  I'm seeing multiple innfeed processes running after INN has been up for more than a day or so.  Seems to happen when I'm reloading configurations such as newsfeeds or incoming.conf.  The new innfeed process starts fine but the old one never seems to die on its own.  It also doesn't stop using standard signals via kill.  I have to kill -9 the process in order for it to quit running.

Here is what it ends up looking like at the process level (if I leave this alone for a few days there will be 4 or 5 innfeed processes running):

news       561  1.4  0.8 172104 34912 ?        Ss   00:58   9:39 /usr/local/news/bin/innd
news      7403  0.3  0.0  12504  3004 ?        S    08:13   0:54  \_ /usr/local/news/bin/innfeed
news     10971  0.1  0.2 148316 11984 ?        S    12:02   0:00  \_ controlchan
news     10972  0.4  0.0  11432  1804 ?        S    12:02   0:00  \_ /usr/local/news/bin/innfeed
news     10973  0.0  0.0   4580  1052 ?        S    12:02   0:00  \_ /usr/local/news/bin/ninpaths -p -d /usr/local/news/log/path/inpaths.%d

When I do a strace on the old innfeed process I see the following:

[news at usenet ~]$ strace -p 7403
Process 7403 attached - interrupt to quit
select(17, [16], [], [], {7, 37802})    = 0 (Timeout)
select(17, [16], [], [], {30, 0})       = 0 (Timeout)
select(17, [16], [], [], {29, 0})       = 1 (in [16], left {21, 446303})
readv(16, [{"503 time out\r\njfrl2le8023d97mmcu"..., 256}], 1) = 14
sendto(3, "<61>Mar  8 12:07:34 innfeed[7403"..., 143, MSG_NOSIGNAL, NULL, 0) = 143
sendto(3, "<61>Mar  8 12:07:34 innfeed[7403"..., 138, MSG_NOSIGNAL, NULL, 0) = 138
close(16)                               = 0
select(0, [], [], [], {21, 0}


When this first started I thought it was happening due to certain configuration parameters that I had tweaked, but I have since reverted those back to defaults with no effect.  I tried hunting around to see if anyone had run into this before but came up empty.  Have you ever seen this behavior?

Regards,
Jesse Rehmer



More information about the inn-workers mailing list