Plesk 10.2 crashing almost hourly

[FiZi]

Bind under Plesk 10.2 crashing almost hourly

Hello,

We're running Plesk 10.2 with the latest patches on CentOS 5.5 32-bit.

It seems the Plesk control panel is randomly crashing on us and taking out Bind (named) with it or vise-versa.

It appears Bind is crashing almost hourly which makes Plesk and our website inaccessible. I completed the 10.2 upgrade 2-3 weeks ago so if it's related it sure took it's sweet time.

This is what I've found in our Plesk logs:

[root@########## log]# cat /var/log/sw-cp-server/error_log
2011-05-09 08:24:34: (log.c.75) server started
2011-05-10 08:08:15: (connections.c.299) SSL: 1 error:1407609C:SSL routines:SSL23_GET_CLIENT_HELLO:http request
2011-05-10 08:08:16: (connections.c.299) SSL: 1 error:140940E5:SSL routines:SSL3_READ_BYTES:ssl handshake failure
2011-05-10 08:44:13: (log.c.135) server stopped
2011-05-10 08:44:34: (log.c.75) server started
2011-05-11 15:24:37: (log.c.75) server started
2011-05-12 10:44:35: (log.c.75) server started
2011-05-12 14:06:12: (connections.c.299) SSL: 1 error:1407609C:SSL routines:SSL23_GET_CLIENT_HELLO:http request
2011-05-12 14:06:12: (connections.c.299) SSL: 1 error:1407609C:SSL routines:SSL23_GET_CLIENT_HELLO:http request
2011-05-12 14:06:33: (connections.c.299) SSL: 1 error:1407609C:SSL routines:SSL23_GET_CLIENT_HELLO:http request
2011-05-12 14:06:33: (connections.c.299) SSL: 1 error:1407609C:SSL routines:SSL23_GET_CLIENT_HELLO:http request
2011-05-12 14:06:41: (connections.c.299) SSL: 1 error:14094418:SSL routines:SSL3_READ_BYTES:tlsv1 alert unknown ca
2011-05-12 14:06:41: (connections.c.299) SSL: 1 error:140940E5:SSL routines:SSL3_READ_BYTES:ssl handshake failure
2011-05-13 10:32:37: (log.c.75) server started
2011-05-13 10:44:30: (connections.c.299) SSL: 1 error:1407609C:SSL routines:SSL23_GET_CLIENT_HELLO:http request
2011-05-13 11:21:22: (log.c.75) server started
2011-05-13 11:21:35: (connections.c.299) SSL: 1 error:1407609C:SSL routines:SSL23_GET_CLIENT_HELLO:http request
rndc: connect failed: 127.0.0.1#953: connection refused
rndc: connect failed: 127.0.0.1#953: connection refused
rndc: connect failed: 127.0.0.1#953: connection refused
2011-05-13 12:07:23: (log.c.75) server started

I checked out those rndc errors and they are a result of Bind crashing. The rndc keys match in the config files.

I found a thread with similar errors on the support forums but the KB articles linked didn't apply to Plesk 10.2 and I checked our secure.log and we're not being attacked.

I'd love to show you some logs from named but it doesn't appear that information is logged by default and when I attempt to configure it log files are not created. There is nothing in /var/log/messages that applies.

Any ideas?

 

[FiZi]

Temporary solution. Add the following cronjob.

*/15 * * * * service named restart

 

[IgorG]

http://kb.odin.com/en/8119

 

[FiZi]

http://kb.odin.com/en/8119

Thanks. I did find that article and we're already running DenyHosts. I don't think SSH brute-force attempts are causing our problem.

The named service (Bind) just seems to crash. When it crashes our DNS goes out and we can't get to Plesk.

I'm sorry I confused the issue with my original post because I hadn't completely figured out what was going on.

How should I go about troubleshooting BIND? I can't find any logs for it or any hints as to why it's just crashing after running for about 30 minutes.

 

[IgorG]

Log file of BIND is /var/log/messages

 

[FiZi]

Log file of BIND is /var/log/messages

Thanks. I did mention I looked in there and couldn't find anything.

I found out how to enable logging for BIND. As root:

rndc trace 7

This will log to a default location of "/var/named/run-root/var/named.run".

Of course now that I've figured this out our DNS server isn't crashing anymore.