Bug #34756 SR/NR can fail with "Error 630 during restore of TX"

• Submitted: 22 Feb 2008 8:42
• Modified: 2 Jun 2008 5:30
• Reporter: Jonas Oreland
• Status: Closed
• Category: MySQL Cluster: Cluster (NDB) storage engine
• Severity: S3 (Non-critical)
• Version: CGE-6.1
• OS: Any
• Assigned to: Jonas Oreland
• CPU Architecture: Any

[22 Feb 2008 8:42] Jonas Oreland

Description:
this is a regression introduced with bug#34033
/me curses loudly, 899 is haunting me

problem is that same PK could get different ROWID

problem could be solved by:
1) handling it in restore.cpp
2) impl. a "new" lcp-snapshot which does not block delete, but prevent
  inserts from entering LCP

How to repeat:
so know way,
found by divine luck while thinking about it

Suggested fix:
2) seems easier and less error prone

[25 Mar 2008 15:42] 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/44410

ChangeSet@1.2535, 2008-03-25 16:42:06+01:00, [e-mail] +5 -0
  ndb - bug#34756 - SR/NR can fail with "Error 630 during restore of TX"
    prevent same PK from getting different ROWID in same LCP

[25 Mar 2008 15:56] Jonas Oreland

pushed to 51-ndb, telco-6.1 (!), 62,63 & 64

[2 Apr 2008 20:20] Jon Stephens

Documented in the 5.1.23-ndb-6.3.11 changelog as follows:

        In some situations, a problem with synchronizing checkpoints between
        nodes could cause a system restart or a node restart to fail with
        the error Error 630 during restore of TX.

Left Patch Queued status pending additional merges.

[4 Apr 2008 22:36] Jon Stephens

Also noted fix in the 5.1.23-ndb-6.2.15 changelog.

[2 Jun 2008 5:30] Jon Stephens

Closed following discussion with Jonas.

[12 Dec 2008 23:27] Bugs System

Pushed into 6.0.6-alpha  (revid:[e-mail]-20080325154206-46773) (version source revid:[e-mail]-20080516085603-30848) (pib:5)