MySQL Bugs: #67883: Invalid LCP (Ndbd file system inconsistency error)

Bug #67883	Invalid LCP (Ndbd file system inconsistency error)
Submitted:	12 Dec 2012 12:30	Modified:	17 Feb 2013 20:00
Reporter:	Davide F	Email Updates:
Status:	No Feedback	Impact on me:	None
Category:	MySQL Cluster: Cluster (NDB) storage engine	Severity:	S2 (Serious)
Version:	mysql-5.5.27 ndb-7.2.8	OS:	Linux (Debian 6.0)
Assigned to:	Assigned Account	CPU Architecture:	Any
Tags:	LCP, ndbd, ndbmtd

Description:
Discalimer: this is a test cluster, even if inserted in a production environment, so please do not answer with best practices: I know them, it's just a test, so please ignore the warning in the error logs.

The error is very simple: when I try to test the MySQL Cluster real HA, by killing the machine where a datanode is running, ndbmtd cannot recover its data using the LCP snapshot it has on disk, I have to use --initial and force the data copy from the other nodes

This happens all the times.

Time: Wednesday 12 December 2012 - 13:08:41
Status: Ndbd file system error, restart node initial
Message: Invalid LCP (Ndbd file system inconsistency error, please report a bug)
Error: 2352
Error data: T24F3
Error object: RESTORE (Line: 1283) 0x00000002
Program: ndbmtd
Pid: 2754 thr: 3
Version: mysql-5.5.27 ndb-7.2.8
Trace: /usr/local/mysql/data/ndb_4_trace.log.9 [t1..t5]

How to repeat:
- install a basic mysql cluster (3 nodes), NoOfReplicas 3
- get the cluster up and running
- insert some data
- force kill one node
- try to start ndbmtd
- watch it failing

Hi Davide,

I could not reproduce the issue using the instructions you provided on mysql-5.5.27 ndb-7.2.8 (and ndb-7.2.7).

Please provide more details on cluster configuration (eg. config.ini, my.cnf), logs (ndb error report), schema, operations you performed, kind of activity on database when you killed data node and restarted it.

Thanks.

No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".