milleruntime commented on issue #2085: URL: https://github.com/apache/accumulo/issues/2085#issuecomment-847047942
There are a number of factors that could affect Tablet migration (tablet size, number of tablets, number of tservers, recovery, configuration). Are you using the default configured Load Balancer (see `table.balancer` property) DefaultLoadBalancer? If so, it will attempt to keep an even number of tablets across all tservers. When a tablet gets migrated from one tserver to another, it will get unloaded and then loaded on the new tserver. So if the tablet has a Write Ahead Log (WAL) associated with it, it will do recovery on the WAL. If you are having issues with WALs (based on the previous comments) then this would hold up loading of the tablets with the problematic WALs. If you have resolved the WAL problems, then you might want to look into using a different load balancer. See this blog post about the GroupBalancer: https://accumulo.apache.org/blog/2015/03/20/balancing-groups-of-tablets.html -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
