Re: Can't lock table error - Magnus Svensson - com.mysql.lists.cluster

6 messages in com.mysql.lists.clusterRe: Can't lock table error

From	Sent On	Attachments
Yannis Tsakiris	28 Sep 2004 00:24
Magnus Svensson	28 Sep 2004 00:49
Yannis Tsakiris	28 Sep 2004 01:51
Yannis Tsakiris	28 Sep 2004 02:52
Magnus Svensson	28 Sep 2004 04:40
Yannis Tsakiris	29 Sep 2004 00:50

Subject:	Re: Can't lock table error
From:	Magnus Svensson (msve...@mysql.com)
Date:	09/28/2004 04:40:33 AM
List:	com.mysql.lists.cluster

Quoting Yannis Tsakiris <gtsa...@dolnet.gr>:

your MySQL Server has for some reason lost the connection to the NDB Cluster storage engine. That is why you get the error "Can't lock file(errno: 4009)". Why does this happen? In order to diagnose the problem you should look in the cluster log file. It can be found in the directory of you MGM server and will be called ndb_<nodeid>_cluster.log. In that file look for messages indicating like "Node <nodeid> disconnected" and "Node <nodeid> missed heartbeat 2" etc. You can also send that and your config.ini file to the list. If the computers are heavily loaded the most likely problem is that the mysqld is excluded from the cluster because it misses too many heartbeats, you should increases the below two parameters in config.ini.

The computers are not that much loaded, they are a pair of pIII at 1 GHz, 2 cpu's on each computer with 1 GB each one. The average load is about 0.25. Anyway, I'll try to increase the interval between the heartbeats and see what will happen.

Also, instead of restarting the NDB nodes have you tried to restart the MySQL Server that get this message? Will that resolve the problem?

I'll try this alternative as soon as it happens again...

I also send you my current config.ini: (2 Computers, 2 MGMs, 2 DBs, 2 APIs, 1 node group, 2 replicas):

[COMPUTER] Id: 1 Hostname: 10.3.90.1

[COMPUTER] Id: 2 HostName: 10.3.90.2

[MGM DEFAULT] PortNumber: 2200

[MGM] Id: 1 ExecuteOnComputer: 1 ArbitrationRank: 1

[MGM] Id: 2 ExecuteOnComputer: 2 ArbitrationRank: 2

[DB DEFAULT] NoOfReplicas: 2 StopOnError: N

[DB] Id: 3 ExecuteOnComputer: 1 FileSystemPath: /usr/local/mysql/cluster/db/

[DB] Id: 4 ExecuteOnComputer: 2 FileSystemPath: /usr/local/mysql/cluster/db/

[API] Id: 5 ExecuteOnComputer: 1

[API] Id: 6 ExecuteOnComputer: 2

[TCP DEFAULT] PortNumber: 28002

[TCP] NodeId1: 1 NodeId2: 3

[TCP] NodeId1: 1 NodeId2: 4

[TCP] NodeId1: 2 NodeId2: 3

[TCP] NodeId1: 2 NodeId2: 4

[TCP] NodeId1: 3 NodeId2: 5

[TCP] NodeId1: 3 NodeId2: 6

[TCP] NodeId1: 4 NodeId2: 5

[TCP] NodeId1: 4 NodeId2: 6

[TCP] NodeId1: 3 NodeId2: 4

You can remove all [TCP] sections in your config.ini files since this will be calculated automatically by the mgm server when it reads the config file(the rule is simple, all nodes must have connections between them). Your problem could be because there is a connection missing! Easisest way is to remove all. :)

Have you looked in the ndb_1_cluster.log?

If the error occurs again you could type the SHOW command in the ndb_mgm and send the output to the list.

Best regards Magnus

---------------------------------------------------------------- This message was sent using IMP, the Internet Messaging Program.