Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
Munin log transfer
Content from Munin Trac Wiki. Was written by janl on 2006-05-28.
This is a RFC for the time being.
Munin-node collects a lot of logs. These reside on the nodes. If they were on the master node, quite possibly presented in the web GUI this might prove helpful. At the very least it would provide a one-stop point for troubleshooting munin. Hopefuly it will give an impetus to plugin writers to detect more errors - somebody gives a hoot and the messages can be seen on the servers.
At the moment this is a very rough sketch just to dump my brain.
Introduce a new protocol command called "nodelog" that works much like the spooling extention with regards to timestamps and storage of logs.
The munin-node needs to catch all STDERR output from each plugin (and it's child processes) and prefix each line to help identify the plugin/hostname and timestamp (much like syslog). Then replay the logged materials from the given time when the server issues a "nodelog $timestamp" command.
The capability called "nodelog" signals the availability of this feature.
If the munin-node applies an appropritate RE on each line of the plugins STDOUT it may also catch non-compliant output there which are in reality error messages.
Special tags such as MUNINERROR: and MUNINWARNING: in the STDERR output could color the graphs appropriately and cause the text following the tag to be included in the index graph page for the host. These should probably be explicitly acknowledged...?
How can this be relayed to nagios etc?