Details
-
Suggestion
-
Resolution: Unresolved
-
None
-
None
-
0
-
1
-
Description
Issue Summary
When launching a new node with using AWS AutoScaling, the new node will be launched without local home files at first. Because of this, the index recovery work in the new node will tries to execute rebuilding index without "/path/to/local/node/home/journal/system_maintenance".
From this work, the new node will starts to execute all `system_maintenance` entries from db, and sometimes index data will get old with old system_maintenance index snapshot if it exists. If the index data in
system_maintenance snapshot is old(system_maintenance record and snapshot is created by rebuilding index), some latest index data will be lost and it affects macro and search result in Confluence.
This is reproducible on Data Center: yes
Steps to Reproduce
- Step 1: Execute re-index, and system_maintenance will be recored in journal_entry table.
- Step 2: ~ long time ~ => The difference between the latest journal entry and re-index data are getting big.
- Step 3: Add a new node without local home via AWS AutoScaling
- Step 4: The new node executes Index Recovery.
- Step 5: However, the node will try to start to execute all `system_maintenance` entries from db until all executed, and old system_maintenance snapshot file in Step 1 will be used, and the index data will be getting old.
Expected Results
Improve rebuilding index process for launching of the new node which has no local home via AWS AutoScaling
Actual Results
The new node will starts to execute all `system_maintenance` entries from db, and always the index data will get old with old index snapshot at last rebuilded one if it exists.
Workaround
Currently there is no known workaround for this behavior. A workaround will be added here when available
To avoid this mechanism, you can add a new node with existing local home from other node, but it is difficult with using AWS AutoScaling.
Note:
Remove old index snapshot file which corresponds old system_maintenance record in Shared-Home, the index data will not be getting old one.
Attachments
Issue Links
- is related to
-
CONFSERVER-81452 Improve the documentation of Index Propagation in Confluence DC cluster, how does index will be propagated
- Gathering Interest
-
CONFSERVER-81501 Improve the Error message and log level of "LuceneIndexSnapshotException: No snapshot found in shared home for index MAIN_INDEX with journal entry id xx." in JournalSystemMaintenanceTaskQueue
- Gathering Interest