Open HariSekhon opened 7 years ago
Yeah, we definitely need that back-off code in there. One of our engineers wrote something for AsyncHBase that we may be able to co-opt for this purpose.
We'll get this in 2.3.1 most likely. It'll need some testing.
👍
I'm trying to bulk import a lot of data from an OpenTSDB migration, and I'm facing this error a lot! As I'm importing a lot of data, I'm not surprised HBase can keep up and is throwing this exception when compatcing and so on. But each time it does this, tsdb import
does not handle it and just exit with error, failing the whole import. This is a pain to manage.
Please handle this exception! I do not need an exponential backoff. A just simple wait and retry loop will do the trick! It will save my whole import and will help manage this big opentsdb migration.
Still need it, just gotta work it out.
When bulk loading to OpenTSDB on HBase I consistently get a RegionTooBusyException. I've checked #757 but we've already solved that bug as I upgraded our clusters to HDP 2.5 which contains the patch for HBase and we've tested the .tmp data volume doesn't increase.
I think the correct response to this would be for the opentsdb import bulk loader to catch and retry with exponential backoff, similar to the solution accepted for #867.
It currently looks like it retries 4 times immediately (timestamps show 1 millisecond apart), which doesn't give enough time for HBase to clear it's backlog.