Charles Connell
Charles Connell
> By default we use AsyncProtobufLogWriter, so we also need to change the code there. Thanks for pointing this out, I've addressed it now
thank you @lukesteensen , I've rebased now
I would say the issue with Kubernetes secrets mounted in containers is a little bit different than the one mentioned in that blog post. However, simply polling the file of...
> Sorry for coming back too late here. Quick question: why have you replaced the WatchService implementation instead of introducing the polling mechanism as a new and configurable option? No...
It's true, WatchService uses a little less CPU time. That's a trade-off.
@Apache9 I've addressed some of your comments, still working on others.
Test failure appears unrelated
@Apache9 Understood. I've brought back the thread pool usage and added a barrier to force onComplete() to wait until onRegionComplete() is done.
> For me I prefer we just send the request to master and let master to trigger reload for region servers. That is exactly what this PR implements. > For...
I understand now. Remote procedures are a better way for the HMaster to trigger something to happen on a RegionServer, but I did not consider them at first. I've re-written...