taklwu commented on PR #6981:
URL: https://github.com/apache/hbase/pull/6981#issuecomment-2873870011

   @Apache9 Hi Duo, thanks for your comment, and let me try to explain below 
why this patch is still needed below, and please see if that makes senses.
   
   > we can load the location in master region and find out that the region 
server is gone and schedule a SCP for it
   
   are you referring we need to use hbck tool to schedule SCP for it? 
   Based on the tests added in this PR—specifically 
TestRecreateCluster#testRecreateCluster_UserTableEnabled_CleanupWALAndZNodes—we've
 confirmed that the current master branch still does not automatically schedule 
SCPs for unknown servers. However, manually triggering SCPs via hbck does 
resolve the issue. 
   
   This PR aligns with the suggestion made by @petersomogyi in [this 
comment](https://github.com/apache/hbase/pull/2114#issuecomment-865087242): we 
could introduce an optional feature flag to automatically schedule SCPs when 
the configuration knob `hbase.master.assign.regions.on.unknown.servers` is 
enabled.
   
   Here the new test 
`testRecreateCluster_UserTableEnabled_CleanupWALAndZNodes_WithRecoverUnknownServer`
 confirmed that only when we enable this feature, the master and cluster could 
start without any further manually operations.
   
   > And in [HBASE-26245](https://issues.apache.org/jira/browse/HBASE-26245), 
we will also store the region server list in master local region, so even if 
the WAL Directories are gone, we could still find the region server list.
   
   that patch solves the problem of region server list, but those are unknown 
servers, the cluster would be hanging and till a manual SCP via HBCK, it won't 
move further. 
   
   --- 
   
   the goal to have less manual operations if possible with a opt-in feature 
(that would be compatible with existing logic), also, this is different from 
the original attempt in PR https://github.com/apache/hbase/pull/2113 that we 
don't [remove/delete meta 
table](https://github.com/apache/hbase/pull/2113/files#diff-45060be7c234e0c21951395df571e80ad7521af333c65dade65796a366b54570R78),
 and only schedule SCP those unknown servers. 
   
   


-- 
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