From: Michael Dykman
Date: July 14 2004 4:37pm
Subject: Re: Crash during data import
List-Archive: http://lists.mysql.com/cluster/129
Message-Id: <40F56163.4010302@dykman.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit

might I suggest chunking such a huge job?

try
$ mysqldump --no-data
to get the table definitions, import those and then
$ mysqldump  --no-create-info [--extended-insert]
for the data..  then split the data file on line boundries (it will be 
insert statements all on one line ) into smaller files.

  - michael dykman
  - michael@stripped


Luca Villani wrote:
> Hi all.
> 
> I'm testing data import for cluster.
> 
> I've done a dump of a working database, with
> 
> 	mysqldump --opt global >global.sql
> 
> resultig file has a size of 2463157827 byte.
> 
> I try to import it onto a working 2node-2api cluster, but I can load only 
> three table, due to this error:
> 
> 
> [root@delfi02 tempo]# /usr/local/mysql/bin/mysql -h delfi01.backplane \
> 
>>global <global.sql
> 
> ERROR 1297 at line 81: Got temporary error 4010 'Node failure caused abort of 
> transaction' from ndbcluster
> [root@delfi02 tempo]
> 
> 
> I can see only this in logfils:
> 
> ==> cluster.log <==
> 2004-07-14 15:24:43 [MgmSrvr] INFO     -- Node 2: Local checkpoint 3 started. 
> Keep GCI = 27 oldest restorable GCI = 27
> 2004-07-14 15:32:12 [MgmSrvr] ALERT    -- Node 1: Node 3 Disconnected
> 2004-07-14 15:32:12 [MgmSrvr] INFO     -- Lost connection to node 3
> 
> ==> error.log <==
> Date/Time: Wednesday 14 July 2004 - 15:32:11
> Type of error: error
> Message: Internal program error (failed ndbrequire)
> Fault ID: 2341
> Problem data: DbtuxCmp.cpp
> Object of reference: DBTUX (Line: 82) 0x0000000a
> ProgramName: NDB Kernel
> ProcessID: 1813
> TraceFile: /usr/local/mysqlNDB_TraceFile_2.trace
> ***EOM***
> 
> Date/Time: Wednesday 14 July 2004 - 15:32:11
> Type of error: error
> Message: No message slogan found
> Fault ID: 0
> Problem data: Signal 11 received
> Object of reference: Main.cpp
> ProgramName: NDB Kernel
> ProcessID: 1813
> TraceFile: /usr/local/mysqlNDB_TraceFile_3.trace
> ***EOM***
> 
> 
> ==> cluster.log <==
> 2004-07-14 15:32:12 [MgmSrvr] ALERT    -- Node 1: Node 2 Disconnected
> 2004-07-14 15:32:12 [MgmSrvr] INFO     -- Lost connection to node 2
> 
> 
> 
> The mysqldump's --opt switch implies fast insert statements syntax in 
> resulting dump file, is this syntax well handled by cluster?
> 
> Have I some another problems?
> 
> 
>