Details
-
Bug
-
Resolution: Fixed
-
Critical
-
5.5.0
-
Untriaged
-
Unknown
Description
1) populate bucket with N items
2) create partition index (create index name on default(name) partition by hash(name))
3) add a new indexer node -> rebalance in
4) remove the newly indexer node -> rebalance out
5) select name from default where name is not missing -> return N rows
6) kill indexer
7) wait till recovery is done then run query again
At step 7, it returns fewer items than step 5.
It seems that the plasma file got reset after recovery.
For example, before killing indexer (but after rebalance is done)
John-couchbase-MacBook-Pro-2:default_name_17977797937491420417_1.index jliang$ cd mainIndex/
|
John-couchbase-MacBook-Pro-2:mainIndex jliang$ ls -lrt
|
total 176
|
-rwxr-xr-x 1 jliang staff 81920 Apr 5 20:33 log.00000000000000.data
|
drwxr-xr-x 3 jliang staff 102 Apr 5 20:33 index
|
-rwxr-xr-x 1 jliang staff 8192 Apr 5 20:33 header.data
|
After killing indexer and restart,
John-couchbase-MacBook-Pro-2:default_name_17977797937491420417_1.index jliang$ cd mainIndex/
|
John-couchbase-MacBook-Pro-2:mainIndex jliang$ ls -lrt
|
total 24
|
-rwxr-xr-x 1 jliang staff 4096 Apr 5 20:37 log.00000000000000.data
|
-rwxr-xr-x 1 jliang staff 8192 Apr 5 20:37 header.data
|
John-couchbase-MacBook-Pro-2:mainIndex jliang$ ls -lrt
|