Bug #46856 Cluster crash during schema transaction can lead to inconsistent state
Submitted: 21 Aug 2009 12:56
Reporter: Jonas Oreland Email Updates:
Status: Verified Impact on me:
Category:MySQL Cluster: Cluster (NDB) storage engine Severity:S3 (Non-critical)
Version:mysql-5.1-telco-7.0 OS:Any
Assigned to: CPU Architecture:Any

[21 Aug 2009 12:56] Jonas Oreland
start schema transaction
create logfile group
create undofile
commit schema trans; // crash

after a system restart, the undofile will remain on disk

How to repeat:
remove #if 0 in testDict -n Bug36072

Suggested fix:
When finding pending transactions during restart,
rebuild transaction state and run abortPrepare work-flow,
or possible a set of new function pointers e.g

- restartAbort
- restartCommit
- restartComplete
[21 Aug 2009 12:58] Jonas Oreland
D2: can in some worst cases cause crash, but that is likely very rare
W3: backup/restore
I4: very slim time-window