DB failover can fail under rare circumstances
Details
| Detail name | Value |
|---|---|
| Changelog Number | 8695 |
| Type | Bug |
| Status | Open |
| Affected Versions | Exasol 7.0.0, Exasol 6.0.16, Exasol 6.2.1, Exasol 6.1.6, Exasol 7.1.0, Exasol 8.0.0 |
Problem
The failsafety mechanism may fail to activate correctly in very specific cases:
During a node failure, it is in some cases possible that the information that the database processes have failed is received before the information that the node itself has failed. In this case the failsafety mechanism will try to restart the database on the crashed node because the information that the node is unavailable has not been received yet.
This causes the database to fail to start, because an "active" node is offline. Because databases that fail to start in the first place are not re-started automatically, this scenario requires manual intervention to re-start the database.
Workaround
- Deactivate the affected node in EXAoperation (if e.g. a hardware failure is suspected and will be investigated and the node should not be used as a reserve node in the mean time)
- Start the database
- Re-Activate the affected node (after investigating e.g. a potential hardware issue)
