Re: Cluster on Solaris 9: Bus Error, Core Dump - Alexander Haubold - com.mysql.lists.cluster

13 messages in com.mysql.lists.clusterRe: Cluster on Solaris 9: Bus Error, ...

From	Sent On	Attachments
Alexander Haubold	20 Jul 2004 12:56
Mark Kim	20 Jul 2004 13:29
Alexander Haubold	22 Jul 2004 12:28
Alexander Haubold	23 Jul 2004 00:30
Magnus Svensson	23 Jul 2004 01:21
Magnus Svensson	23 Jul 2004 01:25
Alexander Haubold	23 Jul 2004 13:06
Alexander Haubold	26 Jul 2004 22:10
Alexander Haubold	27 Jul 2004 11:37
Mark Kim	27 Jul 2004 13:32
Alexander Haubold	27 Jul 2004 14:39
Magnus Svensson	28 Jul 2004 03:11
Alexander Haubold	28 Jul 2004 07:13

Subject:	Re: Cluster on Solaris 9: Bus Error, Core Dump
From:	Alexander Haubold (ah2...@columbia.edu)
Date:	07/28/2004 07:13:53 AM
List:	com.mysql.lists.cluster

Hi Magnus,

Here is the config.ini I am using. It has been decimated to what I believe are the most generic options. I have also removed the "ByteOrder: Big" option under Computer.

Thanks for looking into this.

- Alex

[COMPUTER] Id:1 HostName: icarus

[MGM] Id:1 ExecuteOnComputer: 1 PortNumber: 10000

[DB DEFAULT] NoOfReplicas: 2

[DB] Id: 2 ExecuteOnComputer: 1 FileSystemPath: /usr/local/mysql/ndb/db-1/data

[API DEFAULT] ArbitrationRank: 1

[API] Id: 3 ExecuteOnComputer: 1

At 06:12 AM 7/28/2004, Magnus Svensson wrote:

On Tue, 2004-07-27 at 20:37, Alexander Haubold wrote:

Hi once more,

Here are some other observations related to the core dump:

Starting mysql/libexec/ndbd works (with a reasonable error message and no core dump) when: - there is no Ndb.cfg or - the Ndb.cfg has a wrong server/port entry or - the management server is not running, but the Ndb.cfg entried are correct. When modifying the configuration to match the server and port number, the core dump occurs.

Hi,

your observations are correct. The error occurs when ndbd connects to ndb_mgmd and "sucessfully" retreives it's configuration. While unpacking the received "blob" which contains the configurations for ndbd a bus error occurs and the ndbd process will crash.

We will make this into a bug report and investigate it further.

If you want to try anything, then change your config.ini file to see if removing some of the entries in there will help. :) I have seen that removing the ByteOrder parameter could help.

And maybe you can send us yout config.ini file to help debug the problem?

Best regards Magnus

Here are the 2 configurations I have tried for Ndb.conf:

1. nodeid=1; host=localhost:10000

2. OwnProcessId 2 localhost 10000

The management server is running on port 10000.

Any new ideas?

Thanks.

- Alex

At 04:21 AM 7/23/2004, Magnus Svensson wrote:

On Fri, 2004-07-23 at 09:30, Alexander Haubold wrote:

Hello,

I have now built the latest version of MySQL cluster, but without --with-debug. In my previous post I have mentioned that an error occurs when compiling ndb/src/kernel/vm/Emulator.cpp related to the function "signal". This error occurs only when --with-debug is passed to configure. Without it, the package compiles. However, even with the newest version, the core dump still remains when running ndbd. Perhaps this is related to the "signal" error when using --with-debug?! Can someone suggest a way to get some more debugging information beyond a core dump? Unfortunately, I am unfamilar with strace, and running it produces the output "ERROR: tracer already exists".

Any help would be appreciated.

Hi,

a core dump is the most exact way to disagnose that kind of crash.

Please do: gdb <path>/ndbd core

Then when gdb starts: (gdb) bt

and send the result of the backtrace to the list.

Best regards Magnus

Thanks.

- Alex

At 04:29 PM 7/20/2004, Mark Kim wrote:

How recently did you pull mysql cluster down from bitkeeper? Also, did you configure it with debug (--with-debug) turned on (if you compiled

with

gcc) or

run it with strace? Could you send us the output of either?

Mark

Alexander Haubold wrote:

Hi everyone,

I, too, am having problems getting MySQL Cluster (4.1.4) to run on Sparc/Solaris 9. While the management server starts just fine,

ndbd

appears

to not work. The error is simply: Bus Error (core dumped)

On a side note, typing in the command "SHOW PARAMETERS" on the management server prompt results in the following: _mgmtSrvr.getConfigInfo()->print;

Any suggestions would be welcome.

Thanks.

362 Riverside Dr. #5B3 New York, NY 10025

unsubscribe: http://lists.mysql.com/cluster?unsub=mar...@sgi.com

unsubscribe: http://lists.mysql.com/cluster?unsub=ah2...@columbia.edu

-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+- Alexander Haubold Columbia University - SEAS 2001, 2003

362 Riverside Dr. #5B3 New York, NY 10025

Phone: +212-853-8239 Email: ah2...@columbia.edu WWW: http://www.aquaphoenix.com

-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+- Alexander Haubold Columbia University - SEAS 2001, 2003

362 Riverside Dr. #5B3 New York, NY 10025

Phone: +212-853-8239 Email: ah297 [at] columbia [dot] edu WWW: http://www.aquaphoenix.com

-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+- Alexander Haubold Columbia University - SEAS 2001, 2003

362 Riverside Dr. #5B3 New York, NY 10025

Phone: +212-853-8239 Email: ah297 [at] columbia [dot] edu WWW: http://www.aquaphoenix.com