List:General Discussion« Previous MessageNext Message »
From:Bheemsen Aitha Date:October 27 2012 11:56pm
Subject:How to start a cluster using MCM after crash and prepare for restore
using ndb_restore?
View as plain text  
Hi,

We have MySQL cluster 7.2.7 with the following setup.

1 ndb_mgmd, 1 mysqld on one host
2 ndbmtd on another host

We use MySQL cluster Manager known as MCM (v: 1.1.6) to manage the cluster.
We don't use ndb_mgm client at all.

I wanted to simulate and test the restore using ndb_restore.

In order to test the restore script, here are the steps I followed.

1. Created couple of test tables with data in test database.
2. Took an online backup using "START BACKUP" at ndb_mgm prompt.
3. Shutdown the cluster using mcm.
4. renamed the datadir folders for ndbmtd.
5. started ndb_mgmd using mcm.
6. Tried start the first ndbmtd with --initial option in mcm.
7. start mysqld using mcm
8. do restore using ndb_restore

My step #6 is always stuck and not working.

Here is my exact command for step #6 in mcm.

step 5:

mcm> stop process 49 -f attcluster;
+------------------------------+
| Command result               |
+------------------------------+
| Process stopped successfully |
+------------------------------+
1 row in set (17.47 sec)

step 6:

mcm> start process -i -B -f 1 attcluster;
+--------------------------------------------------------------------------------------+
| Command result
            |
+--------------------------------------------------------------------------------------+
| Operation started successfully. Please check the operation result before
continuing. |
+--------------------------------------------------------------------------------------+
1 row in set (0.21 sec)

mcm> show status -r attcluster;
+--------+----------+------------------+----------+-----------+---------+
| NodeId | Process  | Host             | Status   | Nodegroup | Package |
+--------+----------+------------------+----------+-----------+---------+
| 49     | ndb_mgmd | ut06sandboxdb01  | running  |           | 7.2.7   |
| 50     | mysqld   | ut06sandboxdb01  | stopped  |           | 7.2.7   |
| 1      | ndbmtd   | ut06sandboxdb02  | starting | 0         | 7.2.7   |
| 2      | ndbmtd   | ut06sandboxdb02  | starting | 0         | 7.2.7   |
| 51     | ndbapi   | *ut06sandboxdb01 | added    |           |         |
| 52     | ndbapi   | *ut06sandboxdb01 | added    |           |         |
| 53     | ndbapi   | *ut06sandboxdb01 | added    |           |         |
| 54     | ndbapi   | *ut06sandboxdb01 | added    |           |         |
+--------+----------+------------------+----------+-----------+---------+
8 rows in set (0.05 sec)

I see the following in mcmd.log

[1,ndb_mgmd,0]: 2012-10-27 01:42:28 [MgmtSrvr] INFO     -- Node 1: Initial
start, waiting for 2 to connect,  nodes [ all: 1 and 2 connected: 1
no-wait:  ]

I even tried using --nowait-nodes=2 option, but it is not recognized by mcm.
I can't use nodegroup=65536 and "StartNoNodegroupTimeout" options, as they
require editing of config.ini and mcm does not like that.


My questions is did anyone using mcm simulate restore? If yes, how did you
do it?

Any help is appreciated?

Thanks
-BA

Thread
How to start a cluster using MCM after crash and prepare for restoreusing ndb_restore?Bheemsen Aitha27 Oct