logger stopped

[lodder]

I was using the admin console, when i got this error:

system failure: getting database logger connection error code: 
service:FAILURE Method: GetServiceStatusRequest Details:soap:Receiver

after that i check the logs:

*** logswatch version 3.1.1 (pid:21068) started at Sun Feb 17 18:16:08 CET 2008

DBI connect('database=zimbra_logger;mysql_read_default_file=/opt/zimbra/conf/my.logger.cnf;mysql_socket=/opt/zimbra/logger/db/mysql.sock','zimbra',...) failed: Too many connections at /opt/zimbra/libexec/zmlogger line 174
Can't call method "prepare" on unblessed reference at /opt/zimbra/libexec/zmlogger line 61.

version: Release 5.0.1_GA_1902.DEBIAN4.0 DEBIAN4.0 FOSS edition

can this be solved?

[Bill Brock]

"Server status data not available. To see the server status, logger service must be installed."

I get this message in the Admin UI all the time. Sometimes it works and sometimes it doesn't. This has happened with ver,4.5.6 thru 5.0.2. I can use webmin to view the logs and they are always there and running. But the Admin UI seems to lose the connection. I can't find a rhyme or reason because webmin can always show me the logs.

[bdial]

"Server status data not available. To see the server status, logger service must be installed."

I get this message in the Admin UI all the time. Sometimes it works and sometimes it doesn't. This has happened with ver,4.5.6 thru 5.0.2. I can use webmin to view the logs and they are always there and running. But the Admin UI seems to lose the connection. I can't find a rhyme or reason because webmin can always show me the logs.

I don't think this is related. you're running openSuSE right? I think the problem you're encountering is this:

zimbra runs it's log rotate. it moves the /var/log/zimbra.log to an archive directory, then triies to killall -HUP syslogd so it restarts the syslog and begins a new log. Unfortunately, opensuse uses syslog-ng which doesn't respond to a sighup. Therefor the log never gets recreated, until the suse logrotate script runs that night which runs as root and knows to issue a /sbin/rcsyslog restart after it's done. This gets all the logs (including zimbra.log) going again.

the solution to this is to edit the zimbra logrotate to call /sbin/rcsyslog restart and you'll need to give zimbra sudo power for this

[Mank]

I am a SLES 10 user too and was seeing the problem all to often. You're right. Restarting the syslog works.

I think this should be files as a bug since SLES 10 is a supported OS for Zimbra NE.

[cyberdeath]

I am having the exact same problem with SLES 10.

zimbra runs it's log rotate. it moves the /var/log/zimbra.log to an archive directory, then triies to killall -HUP syslogd so it restarts the syslog and begins a new log. Unfortunately, opensuse uses syslog-ng which doesn't respond to a sighup. Therefor the log never gets recreated, until the suse logrotate script runs that night which runs as root and knows to issue a /sbin/rcsyslog restart after it's done. This gets all the logs (including zimbra.log) going again.

the solution to this is to edit the zimbra logrotate to call /sbin/rcsyslog restart and you'll need to give zimbra sudo power for this

What file am I going to need to modify to make what bdial recommended work? I'm not seeing it clearly enough by looking at my crontab. I looked through several of the files that related to logging but to no avail. Suggestions?

EDIT: After some research...I finally found out that the logrotate file is located under /etc/logrotate.d/zimbra

I modified the file and changed the following line:

killall -HUP syslogd 2> /dev/null || true

to:

sudo /sbin/rcsyslog restart 2> /dev/null || true

[LMStone]

I am having the exact same problem with SLES 10.



What file am I going to need to modify to make what bdial recommended work? I'm not seeing it clearly enough by looking at my crontab. I looked through several of the files that related to logging but to no avail. Suggestions?

EDIT: After some research...I finally found out that the logrotate file is located under /etc/logrotate.d/zimbra

I modified the file and changed the following line:

killall -HUP syslogd 2> /dev/null || true

to:

sudo /sbin/rcsyslog restart 2> /dev/null || true

Just saw this same problem on our one-day old 5.0.4 NE SLES10 install.

Thanks for the code update.

Has anyone filed a bug report on this yet?

All the best,
Mark

[cyberdeath]

Added bug report...I didn't see it otherwise (my apologies if it has already been reported):

Bug 26766 - SLES 10 does not properly restart logger resulting in no statistics

...After some research...I finally found out that the logrotate file is located under /etc/logrotate.d/zimbra

I modified the file and changed the following line:

killall -HUP syslogd 2> /dev/null || true

to:

sudo /sbin/rcsyslog restart 2> /dev/null || true

I also should note I edited the /etc/sudoers file to include this (added line):

%zimbra ALL=NOPASSWD:/sbin/rcsyslog

This is required in order to make the aforementioned solution provided work.

[LMStone]

Thanks for the update regarding /etc/sudoers; I would not have thought about that one off the bat (though I probably should have).

Thanks for filing the bug report too; FWIW I would have been happy to file one, and I've added myself as an interested party to your bug report.

P.S. If you paste the output of zmcontrol -v into your profile for the ZCS version, the Zimbra admins here will be much happier, though with a forum name of "cyberdeath" they may give you less grief than they did me a while back! ;-)

[tonster]

Added bug report...I didn't see it otherwise (my apologies if it has already been reported):

Bug 26766 - SLES 10 does not properly restart logger resulting in no statistics



I also should note I edited the /etc/sudoers file to include this (added line):

%zimbra ALL=NOPASSWD:/sbin/rcsyslog

This is required in order to make the aforementioned solution provided work.

I haven't run opensuse recently, but if they are running syslog-ng (like they are in SLES), probably the best work-around to this (until the bug is fixed), is to modify /etc/logrotate.d/zimbra and modify the following:

Bad:
      killall -HUP syslogd 2> /dev/null || true
Good:
      killall -HUP syslog-ng 2> /dev/null || true

This will eliminate the need to modify sudoers, and be consistent with how things are done throughout the rest of the platforms. My belief is that this is likely to be the change that gets incorporated into the fix for a future update to ZCS.

[LMStone]

In a post above bdial reports that SuSE's syslog doesn't respond to sighup.

Have you tested your suggested way on OpenSUSE and SLES10?

I appreciate you want to have as little distro-specific code in the Zimbra deployment scripts as possible, but as a longstanding SuSE shop our experience has been that one of SuSE's core strengths is scripts. The rc<service> scripts and all the SuSEconfig scripts are just the more common ones.

Seems to me restarting SuSE services the "SuSE way" provides a solid safety net whose benefits outweigh the costs of trying to develop/test/maintain/retest a distro-neutral methodology.

That this bug is still here on the fourth point release of Zimbra 5 to me says this kind of code doesn't get unit tested all that much, so if Zimbra elected to do things the "SuSE way" that would cut down on the testing requirements, again, at the "cost" of a little more distro-specific code.

Just my 2 cents...
</soapbox>

All the best,
Mark