I actually have two separate questions, both aimed at performance tuning my database.
The first is about running slapindex on individual cluster members to preserve high availability. Namely, if each cluster member is brought offline individually for reindexing (as opposed to putting them all in read-only mode) and no new indices are added while any of the nodes are offline, are there any risks posed (if the nodes not currently being indexed are left in read/write mode) to the reindexed node when is brought back into the cluster? Or is this a relatively safe plan?
The second revolves around log size. In my current environment, using the 'stats' loglevel, I get about 600-700 lines per second, and the cumulative size of the logs rotated throughout the day can reach in the area of 25G or more. Even though disk space is cheap, I'd love to rotate logs less frequently and consume fewer CPU cycles when gzipping and rotating said logs, while still having good visibility if I need to troubleshoot a problem. Is there any other loglevel folks can recommend that will still give me adequate information if I need to go back and glean information from the logs, while perhaps being a bit less verbose?
Thanks as always.