Bug #43754 Spurious 723 during createEventOperation in presence of node failure

• Submitted: 19 Mar 2009 16:52
• Modified: 20 Mar 2009 10:08
• Reporter: Jonas Oreland
• Status: Closed
• Category: MySQL Cluster: Cluster (NDB) storage engine
• Severity: S3 (Non-critical)
• OS: Any
• Assigned to: Jonas Oreland
• CPU Architecture: Any

[19 Mar 2009 16:52] Jonas Oreland

Description:
If data node dies during a createEventOperation
there is a slight risk that, the surviving datanode will
send a invalid table-reference back to ndbapi,
causing the createEventOperation to fail with error 723
(no such table)

How to repeat:
test_event -n SubscribeNR

---

theoretically this can happen to a mysqld(with replication) starting,
if a datanodes dies it's setting up the replication.
however, it's quite unlikely.

FYI: test_event -n SubscribeNR runs 25 threads doing createEventOperation
  and kills nodes in loop, still problem only occurs rarely

Suggested fix:
make sure a correct table reference is always sent

[20 Mar 2009 8:32] Bugs System

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/69887

2910 Jonas Oreland	2009-03-20
      ndb - bug#43754 - fix spurious 723 when createEventOperation during node-failure

[20 Mar 2009 8:43] Jonas Oreland

pushed to 6.3/6.4 won't fix in 6.2

[20 Mar 2009 8:43] Bugs System

Pushed into 5.1.32-ndb-6.3.24 (revid:[e-mail]-20090320083152-15e12eu2q9u2xdr0) (version source revid:[e-mail]-20090320083152-15e12eu2q9u2xdr0) (merge vers: 5.1.32-ndb-6.3.24) (pib:6)

[20 Mar 2009 8:44] Bugs System

Pushed into 5.1.32-ndb-7.0.5 (revid:[e-mail]-20090320083640-42c4bviok9ovglwd) (version source revid:[e-mail]-20090320083640-42c4bviok9ovglwd) (merge vers: 5.1.32-ndb-7.0.5) (pib:6)

[20 Mar 2009 8:46] Jonas Oreland

P2 mainly due to autotest failure,
otherwise likely P3

[20 Mar 2009 10:08] Jon Stephens

Documented bugfix in the NDB-6.3.24 and 7.0.5 changelogs as follows:

        If data node failed during an event creation operation, there
        was a slight risk that a surviving data node could send an
        invalid table reference back to NDB, causing the operation to
        fail with a false Error 723 (no such table). This could take
        place when a data node failed as a mysqld process was setting up
        MySQL Cluster Replication.