You are viewing a plain text version of this content. The canonical link for it is here.

Posted to dev@kudu.apache.org by Andrew Wong <aw...@cloudera.com.INVALID> on 2019/08/07 05:00:46 UTC

[DESIGN DOC] Controlled tablet server downtime

Hello Kudu developers!

There seems to be a decent amount of interest in a few features relating to
the controlled downtime of tablet servers. To name a few that I've seen
gathering interest, there is:

   - Tablet server maintenance mode (KUDU-2069)
   - Tablet server replica draining / tablet server decommissioning
   (KUDU-1827, KUDU-2914)
   - Unregistering a tablet server from the master (KUDU-2915)
   - Cluster rolling restart (KUDU-2054)
   - Automatic cluster rebalancing (KUDU-2780)

So, I wanted to start a central document to organize my thoughts on each of
these, in hopes that it might steer these features in a more unified
direction.

Please take a look, if you're interested. I'm open to feedback and
discussion.

https://docs.google.com/document/d/12BZqspGjHvQlc-o8XTDixoRol9Q36WJzXLJ6p15Zhf0/edit?usp=sharing

Thanks!
Andrew Wong