Johan,
Hi David.
I thought I would try to take a grip on this now after vacations and stuff. I think your second approach looks like a good way. Have you done any progress regarding this? I thought I should start by figuring out how logfetch works and how to send the messages to the server. I am also primarily looking to use this for switches and other network equipment that do not have any clients of their own.
Sorry for no reply - been on a long vacation.
I haven't made any more progress on this yet. The client message with just [msgs:*] section could be a problem with viewing the client data at any given time if there is another client message coming in for the host
- it would flip-flop between the regular message with all the other sections ([cpu] [disk] etc), and the one with the [msgs:*] section. I guess I could pull the existing client message and inject the additional [msgs:*] sections...
Anyhow, not likely to be a high priority to implement, but if that changes I'll let you know.
David.
/Johan
*From:*David Baldwin [mailto:david.baldwin at ausport.gov.au] *Sent:* den 10 maj 2011 01:48 *To:* Johan Sjöberg *Cc:* xymon at xymon.com *Subject:* Re: [Xymon] Feature request for log test
Johan,
Hi.
We have a syslog server which receives logs from a number of servers and network devices.
Currently we use the log test in Xymon to check for errors in these logs, and it works fine. But it is a little blunt since all log test end up in the same msgs test. It could also be a problem if we get an error in one log file, and need to ack or disable the test. In this case we would not get any alert if there were errors in one of the other logs.
It would have been nice if you could add a client definition to a logfile test in analysis.conf, and report each log file as the device which is originates from. Or maybe as a separate syslog test to distinguish it from the msgs test.
This way we could also set up individual alerts for the different logged devices.
I could, of course, write a client- or serverside script for this, but I always find it difficult to do good log monitoring scripts and it would be nice to be able to use the logic already in Xymon
What do you think? Would anyone else be interested in this feature? I also have no idea how much work it would be.
I have also been looking at this same problem just recently.
My log structure is keyed by IP address with daily files - e.g. /var/log/rsyslog/IP/messages-YYYMMDD - how the files are set up would need to be accommodated.
I've thought of 2 approaches:
- writing a utility from scratch to examine the log files - however this then requires all the message rules to be reimplemented rather than using analysis.cfg
- writing a utility that uses 'logfetch' (xymon client utility) to grab relevant section of logfile and then send a client message (still need to work out what class and other details to include in the header) on behalf of the device which contains [msgs:/var/log/messages] section for the log file. With this approach, if the client has no other client message reporting we're OK, but if not, I'm not sure if it will cause problems also. Mostly I'd be looking at this for switches/firewalls/etc so no direct client report in place.
David.
-- David Baldwin - Assistant Director, Infrastructure (acting) Information and Communication Technology Services Australian Sports Commission http://ausport.gov.au Tel 02 62147830 Fax 02 62141830 PO Box 176 Belconnen ACT 2616 david.baldwin at ausport.gov.au Leverrier Street Bruce ACT 2617
Keep up to date with what's happening in Australian sport visit http://www.ausport.gov.au