Various consumers of anaconda's logging output would find it handy if we did a better job with what we log. Among these consumers are various customer monitoring tools, anamon, snake, etc.
Work that would be required to do this:
- Get remote logging working in loader.
- Audit our log messages to see what can be cut. The storage code in particular is probably logging too much stuff.
- Standardize messages to be more easily machine-parseable, and probably document this "API" somewhere. Promise not to change it.
- Log both when we enter and leave a dispatcher step.
- Log major points in anaconda: all filesystems mounted, package selection starting, etc. This might provide a better way to know where an error happened.
- Log packages being installed, what kickstart scriptlets are being executed, etc.
- Make sure all programs we execute are going into program.log, and perhaps make sure these other log files get sent out via remote logging as well.
- Investigate if there's a better mechanism to use for remote logging than just syslog. Is there something more easily consumable?
Good thing I filed this bug twice, huh?
*** This bug has been marked as a duplicate of bug 524980 ***