We have several application servers (and different applications) running on Amazon AWS and our main method of logging is a combination of statsd, carbon/graphite, and an alert system based off of the metrics coming into carbon. This has been working great for a lot of different situations but I feel we need something else to complement this type of logging. Something that is not as much for streaming or visualization like graphite, but allows easy lookup by time/date, error type, etc... What are some techniques that make reviewing your logs quick and effective? My initial thought is to UDP all logging messages to a central server and insert them into a database (likely MongoDB). Is this a reasonable approach?
-
This problem has already been solved with logstash and elasticsearch.user7146– user71462014年02月24日 08:22:01 +00:00Commented Feb 24, 2014 at 8:22
1 Answer 1
I think you're on the right track. You could centralize logging to a central server and then have logstash consume the logs and store everything in elastic search. Then, you can use Kibana to search / lookup, etc. It's a good complement to statsD+graphite.
-
Ribald, I'm in the middle of setting this up now. It looks super promising, thanks for the tip!tonyl7126– tonyl71262014年02月24日 17:36:06 +00:00Commented Feb 24, 2014 at 17:36