brianloss commented on issue #1644: URL: https://github.com/apache/accumulo/issues/1644#issuecomment-652389884
> > I'd just be worried about halting the tablet server if it's not a transient problem since it will lead to cascading failure of the whole Accumulo cluster. > > That is a possibility. However if unexpected errors cause minor compactions to hang indefinitely across the cluster, that will lead to another set of problems. If there is something that automatically restarts tservers, halting may help work issues out automatically. Fair point. In a cluster with high load, restarting tablet servers is a fairly disruptive event. Having the tablet be unusable might be more disruptive, though. :) I missed that we're going to retry on the NoClassDefFound, so I agree for unknown errors it's better to halt the tablet server. ---------------------------------------------------------------- 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]
