SolidWallOfCode commented on pull request #7843: URL: https://github.com/apache/trafficserver/pull/7843#issuecomment-852196572
I think I've tracked down all the required changes. Fundamentally the issue was 1. In the non-RR case, the failed nameserver was not marked as down, and so HostDB requests for the same FQDN would pile up without triggering nameserver failover. 1. The failover checks would indicate no need for failover even if the the namserver was marked down. As minor other cleanups * The incorrect use of `schedule_at` was replaced by `schedule_in`. * The down flag is cleared after successfully adding the nameserver connection to `epoll`, not before. * Rate limiting for rescheduling `mainEvent` for retries was added, otherwise the number of scheduled events grew gradually but without apparent bound. -- 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]
