Forum OpenACS Development: Re: Getting rid of the haystack in the error log

Collapse
6: Re: Getting rid of the haystack in the error log (response to 1)
Posted by Mark Aufflick on 10/31/03 12:44 PM
I agree with Christian (and others) that we need more levels of debug - so a server like the one Janine talks about could run at, say, level 5, and a server that i don't care too much about but that has limited disk space (or costs me money the more i use), could run at level 1.

Even if we don't want to modify ns_log, surely we could just write an ad_log that supports the extra levels, and ns_log's just what we want.

If bing in tcl space is a problem, it would surely be a very simple server module?

while we're at it, we could build functionality into it like email me on severity 9, or write different log items to different files.

Just some ideas - but i suspectwe have outgrown ns_log.

Collapse
7: Re: Getting rid of the haystack in the error log (response to 6)
Posted by Randy O'Meara on 11/01/03 07:03 PM
Joel,

Your ideas sound good. But why not, while we're at it, provide an interface to dynamically set the level? No reason to force a server restart to change the logging level.

It would be good to be able to do something like: I don't want to see the profuse logging from an uninteresting package... Now, I don't want see logging from another package... and so on. In this way, you could really zero in on the interesting detail while paring away the fluff from other packages, where package in this instance really implies a particular namespace.

This might require an nsv to hold the "uninteresting" package names (or, maybe even the DB so it remembers across service restarts). The TCL call stack could be used to determine the namespace of the caller?

Randy