MySQL Bugs: #58453: Incorrectly handling of GSN_COMMIT/GSN

Bug #58453	Incorrectly handling of GSN_COMMIT/GSN_COMPLETE during take-over
Submitted:	24 Nov 2010 11:13	Modified:	25 Nov 2010 22:53
Reporter:	Jonas Oreland	Email Updates:
Status:	Closed	Impact on me:	None
Category:	MySQL Cluster: Cluster (NDB) storage engine	Severity:	S3 (Non-critical)
Version:		OS:	Any
Assigned to:	Jonas Oreland	CPU Architecture:	Any

Description:
During TC take-over, the take-over-TC node will ask all LQH for transaction
originating from crashing TC. Using LQH_TRANSREQ/LQH_TRANSCONF

If a COMMIT/COMPLETE came from other node in node-group to a node
that had completed LQH_TRANSREQ, an extra LQH_TRANSCONF could be emitted
leading to either a crash in LocalProxy or DBTC

This bug has been around since forever, and likelyhood for it to happen
has increased somewhat with the introduction of ndbmtd.
(from impossible to impossible :)

How to repeat:
new test prg

Suggested fix:
transactions that has already been sent to DBTC
should not be sent again...

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/124851

3343 Jonas Oreland	2010-11-24
      ndb - bug#58453 fix execCOMMIT/execCOMPLETE wrt execLQH_TRANSREQ

Pushed into mysql-5.1-telco-6.3 5.1.51-ndb-6.3.40 (revid:jonas@mysql.com-20101124120652-7fqswrdhoebj6v0o) (version source revid:jonas@mysql.com-20101124120652-7fqswrdhoebj6v0o) (merge vers: 5.1.51-ndb-6.3.40) (pib:23)

Pushed into mysql-5.1-telco-7.0 5.1.51-ndb-7.0.21 (revid:jonas@mysql.com-20101124121655-qc422lt7id0guhiy) (version source revid:jonas@mysql.com-20101124121655-qc422lt7id0guhiy) (merge vers: 5.1.51-ndb-7.0.21) (pib:23)

pushed to 6.3.40, 7.0.21 and 7.1.10

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/124852

3344 Jonas Oreland	2010-11-24
      ndb - bug#58453 fix execCOMMIT/execCOMPLETE wrt execLQH_TRANSREQ

Documented bugfix in the NDB-6.3.40, 7.0.21, and 7.1.10 changelogs, as follows:

        During a node takeover, it was possible in some circumstances
        for one of the remaining nodes to send an extra transaction
        confirmation (LQH_TRANSCONF) signal to the BDTC kernel block,
        conceivably leading to a crash of the data node taking over as
        the new transaction coordinator.

Closed.