This is a brute force approach and should only done as a last resort.
Sometimes bouncing the master(s) helps – make sure all of stopped, not just
failed over to another instance., but it sounds like you have tried that.
You may be able to force it by manually editing zookeeper nodes using zkCli.sh
You are editing zookeeper so make sure you have the correct table id and path(s)
Add permission to write in zookeeper using:
zkCli> addauth digest accumulo:YOUR_SECRET
Under /accumulo/INSTANCE_ID/tables/TABLE_ID/state (going from memory, could be
fate) you can use examine the state
zkCli> ls /accumulo/INSTANCE_ID/tables/TABLE_ID/
there should be mutlple zk nodes, one of which is state (or maybe fate?)
zkCli> get /accumulo/INSTANCE_ID/tables/TABLE_ID/state
– it should say DELETING.
zkCli> set /accumulo/INSTANCE_ID/tables/TABLE_ID/state ONLINE
This will force the table back online. After all tablets are accounted for in
the monitor you should be able to re-issue the delete command via the accumulo
From: Jeff Kubina [mailto:jeff.kub...@gmail.com]
Sent: Wednesday, October 12, 2016 7:38 PM
Subject: Re: How can I remove or fix a hanging droptable in Accumulo?
Lots of bulk ingest assignments. It's been hung for days and even returned
after a reboot of the master and tservers.
On Wed, Oct 12, 2016 at 6:28 PM, Michael Wall <mjw...@gmail.com
<mailto:mjw...@gmail.com> > wrote:
What is the master doing while the droptable is hung?
On Wed, Oct 12, 2016 at 6:14 PM, Jeff Kubina <jeff.kub...@gmail.com
<mailto:jeff.kub...@gmail.com> > wrote:
How can I remove or fix a hanging droptable in Accumulo?