Is there a limit for the number of nodes in an indexer cluster, above wich we will see a performance degradation? With our current growth rate we will probably come towards 128 indexers, each with 16 PB local storage, in a two site indexer cluster. There will also be a search head cluster with 20 to 30 search heads in tow different sites.
from a clustering perspective, as the number of buckets approaches 1 million + we start hitting some performance issues.
having said that, we've scaled to 1k+ indexers, and several million buckets+ recently with Splunk 6.6. there are some recommended timeouts we advocate raising for larger clusters (support should be able to help you out with that!)
Thanks for your reply. We are aware of the 1 million bucket limit, we already hit it and had to upgrade to 6.6 to overcome it