couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From GMail <hvandenb...@gmail.com>
Subject Re: The perfect logger
Date Thu, 08 Dec 2011 20:05:45 GMT
How about some graylog2 GELF support or log stash?

On Dec 8, 2011, at 12:44 PM, Rogutės Sparnuotos wrote:

> Jason Smith (2011-12-07 09:28):
>> Hi, all. I am brainstorming features for the perfect CouchDB logging
>> support. I want to know, if God snapped his fingers and logging in
>> CouchDB was perfect according to You, what would that look like?
>> 
>> I posted a similar email in the development list, but here I am
>> focusing on features that sysadmins and application developers want.
>> 
>> This is the brainstorming and requirements gathering phase. I will
>> compile feedback into a spec on the wiki.
>> 
>> For me, here is what I would like to see, in no particular order:
>> 
>> * Opt-in. No surprising changes to the log format or anything.
>> 
>> * Traditional log targets such as syslog, syslog-ng
>> 
>> * One message per line, no more crazy multi-line stack traces. You
>> should be able to do useful things with `perl -n`
>> 
>> * Javascript errors make more sense. (I know that is vague, it's not a
>> personal pain point but I believe it is problematic for most people.)
>> 
>> * Ability to send debug, info, and error logs to different places (or no place)
>> 
>> * Ability to send Javascript errors and logs to their own place
>> 
>> * Log to a database. This is the elephant in the room. This is huge
>> goal, with lots of complications. It will probably be cut from the
>> first iteration. But this is basically my end game for all this. We
>> want a database or databases which catches requests to our couchapp,
>> vhost rules applied, rewrite rules applied, our log() calls from
>> Javascript, and of course exceptions. And we want a web Couch app to
>> present that and let us sort and filter. And the app will follow the
>> _changes feed and give us a real-time "tail -f" of our work, minus
>> Erlang stack traces.
> 
> Developers want human readable log messages, as Bade Iriabho wrote in his
> thread. Is it at all possible to have those Erlang "crashes" be presented
> in a saner way?
> 
> All the other points sound very nice for sysadmins, unfortunately I have
> nothing to add, because I am still developing.
> 
> -- 
> --  Rogutės Sparnuotos


Mime
View raw message