kabhishek4 opened a new pull request, #5605:
URL: https://github.com/apache/hbase/pull/5605

   Problem : 
   
   
   Rolling restart triggers region movement on the cluster while the 
RegionServers are getting restarted. During this time, the temporary 
RegionServers will start prefetching these files which are only hosted until 
the source RegionServer is restarted. Hence, in this timing window, fetches are 
executed on temporary region servers which takes few minutes. 
   
   Currrently, this behaviour is governed by hbase.hfile.prefetch.delay
   
   Also, file-based block cache gives performance improvement on HBase clusters 
using object storage. COD uses a large ephemeral disk for this purpose.
   
   
   Propopsed Solution :
   
   The number of prefetches on the cluster caused by these temporary region 
movements during Rolling Upgrade (rolling restart) can be mininized by 
introducing delay before the prefetch starts.
   
   To achieve this, making the hbase.hfile.prefetch.delay dynamically 
configurable. The interval will let the rolling restart complete. Resultantly, 
prefetches on temporary region server can be minimized. 


-- 
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.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to