[
https://issues.apache.org/jira/browse/HBASE-17304?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Aman Poonia updated HBASE-17304:
--------------------------------
Description:
While using region_mover.rb in our testing we faced an issue where
region_mover was taking a lot of time to unload regions.
For instance we take 10 region server and create draining znode for 8. So while
moving the regions region_mover.rb doesn't take draining znode in consideration
and it tries to move the region in all the region server which are available so
it fails at the server side and then it sleeps for some second and retries 5
time for each region and iterates over all the region servers. This makes this
process really slow in extreme cases as mentioned above where a lot of nodes in
a cluster are in draining mode.
We need to get the list of draining region servers so that we can avoid them
while moving regions.
was:
While using region_mover.rb in our testing we faced an issue where
region_mover was taking a lot of time to unload regions.
For instance we take 10 region server and create draining znode for 8. So while
moving the regions region_mover.rb doesn't take draining znode in consideration
and it tries to move the region in all the region server which are available so
it fails at the server side and then it sleeps for some second and retries 5
time for each region and iterates over all the region servers. This makes this
process really slow in extreme cases as mentioned above where a lot of nodes in
a cluster are in draining mode.
We need to get the list of draining region servers so that we can avoid them
while moving region.
> Avoid draining region servers in draining mode while moving the regions at
> client side i.e. in region_mover.rb
> --------------------------------------------------------------------------------------------------------------
>
> Key: HBASE-17304
> URL: https://issues.apache.org/jira/browse/HBASE-17304
> Project: HBase
> Issue Type: Improvement
> Components: Client
> Reporter: Aman Poonia
>
> While using region_mover.rb in our testing we faced an issue where
> region_mover was taking a lot of time to unload regions.
> For instance we take 10 region server and create draining znode for 8. So
> while moving the regions region_mover.rb doesn't take draining znode in
> consideration and it tries to move the region in all the region server which
> are available so it fails at the server side and then it sleeps for some
> second and retries 5 time for each region and iterates over all the region
> servers. This makes this process really slow in extreme cases as mentioned
> above where a lot of nodes in a cluster are in draining mode.
> We need to get the list of draining region servers so that we can avoid them
> while moving regions.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
