When using Shared Nothing persistence feature in TIBCO BusinessEvents 5.1.4 (BE), there are possible recovery issues and data inconsistencies when adding/removing/reordering fields in a concept.
book
Article ID: KB0103002
calendar_today
Updated On:
Products
Versions
TIBCO BusinessEvents Enterprise Edition
-
Not Applicable
-
Description
Description: Description ----------------
When you add, remove or reorder fields in a concept and redeploy the code and try to recover from an existing shared nothing store the cache nodes core dump and the recovery fails in some cases. This issue is seen in TIBCO BusinessEvents (BE) 5.1.4 and 5.1.4 hotfix 1.
Symptoms ---------------
Cache engines dumps core and exits, during recovery of spaces from Shared Nothing store. Which leads to complete data loss.
Cause ----------
This is identified a defect introduced in Active Spaces 2.1.4 hotfix2 and 2.1.4 hotfix 4 which is bundled with TIBCO BusinessEvents 5.1.4 and 5.1.4 Hotfix1.
1. Avoid making changes to the concept definition if you want to recover any of the old persisted data.
2. If you need the project to run with the modified concepts and not worry about the existing data, then you can clean up the existing Shared nothing (SN) files from all cache nodes (after a backup of the SN files and the old BE project), and restart the entire cluster from scratch.
3. This issue will be fixed in upcoming BE releases, which will be disclosed to customers via a separate LBN.
Issue/Introduction
When using Shared Nothing persistence feature in TIBCO BusinessEvents 5.1.4 (BE), there are possible recovery issues and data inconsistencies when adding/removing/reordering fields in a concept.
Environment
Operating Systems :
All
TIBCO Product Versions:TIBCO BusinessEvents 5.1.4TIBCO BusinessEvents 5.1.4 HotfFix1
Resolution
For now the resolution or workaround is to avoid making changes to the concept definition when you use inheritence relationship (if you want to recover any of the old persisted data) or clean up the existing Shared nothing (SN) files from all cache nodes (after a backup of the SN files and the old BE project), and restart the entire cluster from scratch.
This defect does not affect the changes made to a regular concept definition.