[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
sigpipes on solaris 8, reprise...
From: |
Marco Marongiu |
Subject: |
sigpipes on solaris 8, reprise... |
Date: |
Thu, 10 Mar 2005 16:04:51 +0100 |
User-agent: |
Debian Thunderbird 1.0 (X11/20050116) |
Hello everybody
First of all, I would like to thank those that replied "offline" to my
previous post and tried to help me solving the problem. Unfortunately,
nothing has changed here :-(((((
Summing up, this is what happens.
I compiled cfengine 2.1.13 (and also tried a CVS snapshot) with both gcc
3.4.2 and 3.0.4, against libpcre 5.0, OpenSSL and Berkeley db 4.3.27 and
3.2.9+patches, on Solaris 8:
# uname -a
SunOS web-pre-se 5.8 Generic_117350-18 sun4u sparc SUNW,Serverblade1
The application sigpipes and dies when it does a remote copy operation,
with the message:
*********************************************************************
Main Tree Sched: copy pass 1 @ Thu Mar 10 15:41:29 2005
*********************************************************************
Checking copy from ims-pre-se:/var/cfengine/repository/config/ntp.conf to
/etc/inet/ntp.conf
Connect to ims-pre-se = 10.39.17.16, port =5308
Updating last-seen time for ims-pre-se
Loaded /var/cfengine/ppkeys/root-10.39.17.16.pub
cfengine:web-pre-se: Received signal 13 (SIGPIPE) while doing [no_active_lock]
cfengine:web-pre-se: Logical start time Thu Mar 10 15:41:29 2005
cfengine:web-pre-se: This sub-task started really at Thu Mar 10 15:41:29 2005
An output of truss -o /tmp/cfagent.truss -f /var/cfengine/bin/cfagent is
attached.
Besides, trying to see if I could solve the problem, I spotted a problem
pattern.
If I do these, something happens (well, half of something).
1) stop "everything cf" running on both the policy host and client:
pkill cf
2) remove the cf_lastseen.db on both the policy host and client:
rm /var/cfengine/cf_lastseen.db
3) run cfagent on the policy host and let all the needed services come up
/var/cfengine/bin/cfagent -v
4) run cfagent on the client:
/var/cfengine/bin/cfagent -v | tee /tmp/cfagent.log
the copy from update.conf works, the copy from cfagent.conf and all
subsequent attempts to run a copy from cfagent die.
If you run cfagent again this is what you see on the server side in the
/var/adm/messages:
Mar 10 16:00:10 ims-pre-se cfservd[19111]: [ID 823470 daemon.error] Denying
repeated connection from ::ffff:10.39.17.40
So it looks like the policy host closes the connection while the client
tries to read or write to it, and cfagent dies sigpipe...
Can anybody help?
Ciao and thanks
--Marco
cfagent.truss.gz
Description: application/gzip
- sigpipes on solaris 8, reprise...,
Marco Marongiu <=