It happened again (the problem with the "can't lock table" errors).
Here is the output of the ndb_mgm program when the cluster is in this
-- NDB Cluster -- Management Client --
Connecting to Management Server: 10.3.90.2:2200
[ndbd(NDB)] 2 node(s)
id=3 (Version: 3.5.1, Nodegroup: 0, Master)
id=4 @10.3.90.2 (Version: 3.5.1, Nodegroup: 0)
[ndb_mgmd(MGM)] 2 node(s)
id=1 (Version: 3.5.1)
id=2 @127.0.0.1 (Version: 3.5.1)
[mysqld(API)] 2 node(s)
id=5 (Version: 3.5.1)
id=6 @10.3.90.2 (Version: 3.5.1)
NDB> all status
Node 3: started (Version 3.5.1)
Node 4: started (Version 3.5.1)
I dont see something strange, though? Do you?
I also checked again the mgm logs and I can't see something like lost of
connection between db and api. This time I restarted one of the two MySQL
servers (notice that all nodes was up and running, no process stops, the
tables are just not available), and I saw that the problem was resolved. So
the conclusion is that when the ndb cluster stops giving access to the
table, I have to restart any of the DB or API nodes.
I want the ndb cluster fully functional all of the time. The only solution I
can think at this moment is two have a crontab job that will restart the
nodes when it cannot access the tables. I believe this is a bad solution.
Please help me with a better one...
----- Original Message -----
From: "Magnus Svensson" <msvensson@stripped>
To: "Yannis Tsakiris" <gtsakiris@stripped>
Sent: Tuesday, September 28, 2004 2:40 PM
Subject: Re: Can't lock table error
> > > Also, instead of restarting the NDB nodes have you tried to restart
> > MySQL
> > > Server that get this message? Will that resolve the problem?