On 7/8/2018 8:46 PM, Goh, Qing Hao wrote:
Can I kindly check what is the best practice for housekeeping archival logs for Solr & Zookeeper? What I meant here is the best practices recommended by Apache and to ensure the middleware is kept healthy?
I have no idea what options ZooKeeper has for log maintenance. It is a separate project, you're on a Solr mailing list.
If you're running the ZK server that's embedded into Solr, then whatever ZK logs will go to the solr log, it will not be sent to its own log. The embedded ZK server is not recommended for production.
The current version of Solr (7.4.0) handles all rotation and deletion of the main solr log with log4j config. For the GC log, this is handled by Java itself. I'm not entirely sure about the console log, but I think that this is overwritten each time Solr starts. Solr will typically silence logs sent to the console once it gets up and running.
If you want to keep logs longer than the current configuration specifies, then you will need to change the logging configuration, possibly to send the logs to a different piece of software.
The current version of Solr is 6.1.0 and Zookeeper is 3.4.6. Thanks in advance
Solr 6.1.0 handles normal log rotation and deletion with log4j config (1.x), but there is a log archival step at startup that is handled in the start script. This involves copying some of the old logs to an archive directory. I'd have to look at the script to discover the precise details. This log archival in the start script was removed in 7.4.0 with the upgrade to log4j 2.x, because log4j 2.x can do log rotation at startup.
Thanks, Shawn
