sylog-ng hangs after upgrade to 1.9.9
Hi, just did an apt-get upgrade today and installed syslog-ng 1.9.9 after this I was fighting some hours with my system. syslog-ng logged a few minutes then it hanged, all logfiles showing the same timestamp. This also caused my mail system ( postfix ) to hang and also caused other problems. After a restart of syslog-ng the problem was gone, also 2 times sylog-ng started logging again after about 15-20 minutes for a short while. I found something on a old mailinglist about 2 sources from the same unix-stream which was the case in my logfile. I removed the second one but the problem still occured. I have also read something about klog runing, which isn't the case on my machine. I have now downgraded syslog-ng to 1.6.9 and it's working fine again. My system is a Debian linux Linux 2.4.31-grsec #21 SMP Fri Dec 2 08:49:48 CET 2005 i686 athlon-4 i386 GNU/Linux Best regards, Cristian Livadaru
On Tue, 2006-03-14 at 18:09 +0100, Cristian Livadaru wrote:
Hi, just did an apt-get upgrade today and installed syslog-ng 1.9.9 after this I was fighting some hours with my system. syslog-ng logged a few minutes then it hanged, all logfiles showing the same timestamp. This also caused my mail system ( postfix ) to hang and also caused other problems.
Hm... could you also quote the _exact_ Debian version? Independent of this I would need some more information what caused the hang. An strace dump of the waiting syslog-ng process or something. -- Bazsi
I had this same problem a week or two ago and had to downgrade. Two separate systems with very similar configs both showed this behavior. A few messages were logged, then after fifteen minutes or less, no more output. The process was running but nobody was home. I regret that I didn't bug it, but I just don't have time on the side to deal with it right now. The system is Debian GNU/Linux, Unstable, on i386. Balazs Scheidler wrote:
On Tue, 2006-03-14 at 18:09 +0100, Cristian Livadaru wrote:
Hi, just did an apt-get upgrade today and installed syslog-ng 1.9.9 after this I was fighting some hours with my system. syslog-ng logged a few minutes then it hanged, all logfiles showing the same timestamp. This also caused my mail system ( postfix ) to hang and also caused other problems.
Hm... could you also quote the _exact_ Debian version? Independent of this I would need some more information what caused the hang. An strace dump of the waiting syslog-ng process or something.
-- # Jesse Molina # Mail = jesse@opendreams.net # Page = page-jesse@opendreams.net # Cell = 1.602.323.7608 # Web = http://www.opendreams.net/jesse/
Quoting Jesse Molina <jesse@opendreams.net>:
I had this same problem a week or two ago and had to downgrade. Two separate systems with very similar configs both showed this behavior. A few messages were logged, then after fifteen minutes or less, no more output. The process was running but nobody was home.
I regret that I didn't bug it, but I just don't have time on the side to deal with it right now.
The system is Debian GNU/Linux, Unstable, on i386.
same for me Debian unstable I can't risc to put back the 1.9.9 version I can just hope it is reproduceable on my testserver which has been upgraded today. I only have a strace of mailq which then hangs while trying to access /dev/log want me to send this? else if the error occours on the testserver I can do watever trace or test you need.
On Tue, 2006-03-14 at 20:22 +0100, Cristian Livadaru wrote:
Quoting Jesse Molina <jesse@opendreams.net>:
I had this same problem a week or two ago and had to downgrade. Two separate systems with very similar configs both showed this behavior. A few messages were logged, then after fifteen minutes or less, no more output. The process was running but nobody was home.
I regret that I didn't bug it, but I just don't have time on the side to deal with it right now.
The system is Debian GNU/Linux, Unstable, on i386.
same for me Debian unstable I can't risc to put back the 1.9.9 version I can just hope it is reproduceable on my testserver which has been upgraded today. I only have a strace of mailq which then hangs while trying to access /dev/log want me to send this? else if the error occours on the testserver I can do watever trace or test you need.
The strace on mailq is useless, I would need the info on syslog-ng itself. Your testing is greatly appreciated. -- Bazsi
Quoting Balazs Scheidler <bazsi@balabit.hu>:
On Tue, 2006-03-14 at 20:22 +0100, Cristian Livadaru wrote:
Quoting Jesse Molina <jesse@opendreams.net>:
I had this same problem a week or two ago and had to downgrade. Two separate systems with very similar configs both showed this behavior. A few messages were logged, then after fifteen minutes or less, no more output. The process was running but nobody was home.
I regret that I didn't bug it, but I just don't have time on the side to deal with it right now.
The system is Debian GNU/Linux, Unstable, on i386.
same for me Debian unstable I can't risc to put back the 1.9.9 version I can just hope it is reproduceable on my testserver which has been upgraded today. I only have a strace of mailq which then hangs while trying to access /dev/log want me to send this? else if the error occours on the testserver I can do watever trace or test you need.
The strace on mailq is useless, I would need the info on syslog-ng itself.
Your testing is greatly appreciated.
I will keep an eye on my testsystem if this happenes there. if not maybee I will install on a weekend where there is not much going on on the production system the version 1.9.9 again and I can try to test it. just give some information on what you need and how I could obtain it. Cris
On Wed, 2006-03-15 at 10:30 +0100, Cristian Livadaru wrote:
I had this same problem a week or two ago and had to downgrade. Two separate systems with very similar configs both showed this behavior. A few messages were logged, then after fifteen minutes or less, no more output. The process was running but nobody was home.
I regret that I didn't bug it, but I just don't have time on the side to deal with it right now.
The system is Debian GNU/Linux, Unstable, on i386.
same for me Debian unstable I can't risc to put back the 1.9.9 version I can just hope it is reproduceable on my testserver which has been upgraded today. I only have a strace of mailq which then hangs while trying to access /dev/log want me to send this? else if the error occours on the testserver I can do watever trace or test you need.
The strace on mailq is useless, I would need the info on syslog-ng itself.
Your testing is greatly appreciated.
I will keep an eye on my testsystem if this happenes there. if not maybee I will install on a weekend where there is not much going on on the production system the version 1.9.9 again and I can try to test it. just give some information on what you need and how I could obtain it.
* an lsof -p <syslog-ng pid>, strace of the hanged process should be enough, * enabling debug symbols, attaching to the process using gdb (gdb -p <pid>) and doing a backtrace (bt command in gdb) would be better, * enabling core dumps before starting syslog-ng, sending a SIGQUIT to it, and sending me the binaries and the core files would be ideal -- Bazsi
participants (3)
-
Balazs Scheidler
-
Cristian Livadaru
-
Jesse Molina