List:Cluster« Previous MessageNext Message »
From:Magnus Svensson Date:August 31 2004 9:08am
Subject:Re: API node crashes after a minute
View as plain text  
On Tue, 2004-08-31 at 09:21, Chad Martin wrote:
> I'm attempting to get a 3 machine/3 node cluster running on Opteron 
> machines running Gentoo Linux.  The problem is that when I run the 
> mysqld process, it dies after a minute with the following in the error log:
> 
> 040831 01:55:49  mysqld started
> 040831  1:56:50  Can't init databases
> 040831  1:56:50  Aborting
>  

Hi,

the mysqld shutdowns because your mysqld can't connect to the cluster,
there is a 60 seconds timeout waiting for the cluster to start. I
recommend that you first start ndb_mgmd and the two(!) ndbd's. Then you
can start your mysqld's.

"BUG#3661 mysqld requires ndbcluster to be started" has previously been
filed for this problem, it will soon be pushed.

> 
> 040831  1:56:50  /usr/local/mysql/libexec/mysqld: Shutdown complete
>  
> 
> 040831 01:56:51  mysqld ended
> 
> I'm running mysql 4.1.4 beta, using the nightly build from 8/3/2004. 
> The compilation went OK, I installed the grant tables using 
> mysql_install_db --user=mysql.  I didn't bother with changing the 
> passwords, since this cluster is on a safe network and will hold nothing 
> but test data.  MySQL resides in /usr/local/mysql.  I set up, on the 
> first machine, /usr/local/mysql/ndb/ with subdirectories of mgmt and 
> ndb1.  I'm using the standard my-medium.cnf with ndbcluster added in the 
> mysqld section.  The plan is to have one management node running on 
> machine 1, and an API and storage node running on each computer.


> 
> I invoke mysqld_safe with --skip-bdb --skip-innodb --user=mysql since I 
> don't need either of those engines.  I've tried using the -i flag with 
> ndbd to no effect.  The management node claims that the storage node 
> never gets past Phase 2 of the startup.  I'm not sure if that's normal.
> 
> I haven't tried getting the other machines to work yet.  When I start 
> the MySQL daemon with --skip-ndbcluster, it works fine.
> 
> I've included the configuration files below, with the commented lines 
> removed.  If anybody can help me with this, I'd really appreciate it. 
> It would mean that the money I spent on these Opteron machines wasn't 
> wasted.
> 
> Thanks,
> Chad Martin
> Arete Studios
> 
> /etc/profile:
> MYSQL_HOME=/usr/local/mysql
> export NDB_HOME=$MYSQL_HOME/ndb
> export PATH="$PATH:$MYSQL_HOME/bin:$MYSQL_HOME/libexec"
> export LD_LIBRARY_PATH=$MYSQL_HOME/lib
> export MANPATH=$MANPATH:$MYSQL_HOME/man
> 
> /usr/local/mysql/ndb/mgmt/config.ini:
> [DB DEFAULT]
> NoOfReplicas: 2
> 
> [COMPUTER]
> Id: 1
> HostName: opteron1
>  
> 
> [COMPUTER]
> Id: 2
> HostName: opteron2
>  
> 
> [COMPUTER]
> Id: 3
> HostName: opteron3
>  
> 
> [DB]
> Id: 1
> ExecuteOnComputer: 1
> FileSystemPath: /usr/local/mysql/ndb/ndb1

Remove the above DB node - according to Mikaels instruction. 

You can still run a mysqld on computer 1, the ndb_mgmd will not require
much processing power.

Best regards
Magnus Svensson


-- 
Magnus Svensson, Software Engineer
MySQL AB, www.mysql.com
Office: +46 709 164 491

Thread
API node crashes after a minuteChad Martin31 Aug
  • Re: API node crashes after a minuteMikael Ronström31 Aug
    • Re: API node crashes after a minuteChad Martin1 Sep
  • Re: API node crashes after a minuteMagnus Svensson31 Aug
    • Re: API node crashes after a minuteChad Martin1 Sep