Details
-
Improvement
-
Resolution: Fixed
-
Major
-
5.5.4
Description
At present, the indexer logs stats as a JSON object within a normal log line. This becomes impenetrable, as the number of indexes on this cluster grows. I realise different people have different workflows and use different tools, but taking less as an example of a reasonably robust way of viewing logs essentially makes scrolling through the logs impossible without first removing these lines.
Taking the sizings, we can see that 325MB of this file (~85%) is these stats blocks:
$ grep 'PeriodicStats' ns_server.indexer.log > indexer_stats.log; ls -lh indexer_stats.log ns_server.indexer.log
|
-rw-r--r-- 1 jflath staff 325M 22 May 16:08 indexer_stats.log
|
-rw-rw---- 1 jflath staff 380M 22 May 12:49 ns_server.indexer.log
|
Don't get me wrong, I love stats, but this is not a suitable or supportable way to log them as the index count increases. The logging guidelines strongly recommend that stats aren't dumped into the "actual" logs for this reason.