HBASE-26864 SplitTableRegionProcedure calls openParentRegions() at a …#4261
HBASE-26864 SplitTableRegionProcedure calls openParentRegions() at a …#4261apurtell merged 2 commits intoapache:masterfrom
Conversation
…wrong state during rollback
|
🎊 +1 overall
This message was automatically generated. |
|
🎊 +1 overall
This message was automatically generated. |
|
🎊 +1 overall
This message was automatically generated. |
|
Nice find! Able to write a unit-test to cover this? |
Apache9
left a comment
There was a problem hiding this comment.
LGTM.
And how could we get IOException when creating procedures for unassign regions?
In our specific case, there is region replica configured for the table. When the primary region is opened at one region server and it needs to split, it sends a SplitRegionRequest to master. At the master side, it checks all replica regions. If one replica region is in RIT (being opened), it will throw out IOException in creating procedures for unassign regions. |
I spent quite some time trying to come up with a unit-test. It is hard to trigger some exception in SplitTableRegionProcedure's state machine. Let me try with regionReplica and a coprocessor to delay open of a replica region. |
14ede35 to
3d1f46b
Compare
|
Created a unitest case. During the test, I found that there is an issue where ignoreRIT flag does not apply to replicaRegions. However, from our production log, I do not find any issue I run into with the unitest. I made sure that with the fix, the new unittest case passes and without the fix, it will fail. @Apache9 and @joshelser, can you review the new patch? Thanks. |
|
🎊 +1 overall
This message was automatically generated. |
|
💔 -1 overall
This message was automatically generated. |
|
💔 -1 overall
This message was automatically generated. |
|
🎊 +1 overall
This message was automatically generated. |
|
🎊 +1 overall
This message was automatically generated. |
|
I run locally the failed test TestQuotaAdmin, it passed. |
|
Ping @joshelser and @Apache9, any comments about the new diff? Thanks. |
#4261) Signed-off-by: Duo Zhang <zhangduo@apache.org> Signed-off-by: Xiaolin Ha <haxiaolin@apache.org>
#4261) Signed-off-by: Duo Zhang <zhangduo@apache.org> Signed-off-by: Xiaolin Ha <haxiaolin@apache.org>
#4261) Signed-off-by: Duo Zhang <zhangduo@apache.org> Signed-off-by: Xiaolin Ha <haxiaolin@apache.org>
apache#4261) Signed-off-by: Duo Zhang <zhangduo@apache.org> Signed-off-by: Xiaolin Ha <haxiaolin@apache.org> (cherry picked from commit 34cb2ee) Change-Id: I03c1df63352f07b679b10edb892cfa749d8b6708
…wrong state during rollback