Re: Crash on backup - Alex Davies - com.mysql.lists.cluster

11 messages in com.mysql.lists.clusterRe: Crash on backup

From	Sent On	Attachments
Troy Davis	25 Aug 2005 07:29
Stewart Smith	25 Aug 2005 21:20
Troy Davis	26 Aug 2005 06:51
Alex Davies	27 Aug 2005 10:24
Troy Davis	29 Aug 2005 07:10
Alex Davies	29 Aug 2005 09:23
Max Mether	29 Aug 2005 09:41
Alex Davies	29 Aug 2005 11:34
Troy Davis	19 Sep 2005 14:33
Alex Davies	19 Sep 2005 16:57
Max Mether	20 Sep 2005 01:36

Subject:	Re: Crash on backup
From:	Alex Davies (davi...@gmail.com)
Date:	08/29/2005 11:34:27 AM
List:	com.mysql.lists.cluster

Dear Max,

You could actually switch steps 4 and 2 to get less downtime. All in all we thus get:

Sorry, being thick! Of course you can change config.ini before you shut down the managment daemon.

Glad to see that everything else was correct, though. Thanks for clearing up the use of --initial.

Troy - please feel free to ask if you still are having problems!

With best wishes,

Alex

1. "start backup" on the management server without creating or deleting any tables while it runs. 2. Change configuration settings on management server as needed. 3. Shut down the management server. 4. Restart management server. 5. Shut down and restart ndbd nodes in a cyclic restart (can be done from the management server with "2 restart" etc. --initial can be obtained with "2 restart -i")

Shut down ONE storage node (ndbd daemon), then start it up again with --initial. Make sure you WAIT for it to fully start. Then shut the next storage node (ndbd daemon) down, start it up with --initial. Repeat for all storage nodes.

You don't *have* to do --initial unless you change parameters that affect the disk storage of the nodes (REDO logs etc.).

I can't think of any reason for you to need to restart your SQL nodes as part of an rolling upgrade or a online config change but someone could correct me on this (or anything else, for that matter!)

No, no reason. If (for some reason) the SQL nodes fail to connect after the rolling restarts, then restarting the SQL nodes should do the trick.

This will also work for online upgrades (though not from 4.1 to 5 btw) if you simply upgrade the managment server rather than change the config and in between shutting down and starting with --initial each storage node you upgrade them.

Yes, this should work. As said not between 4.1 and 5.0 but between 5.0.9 and 5.0.11 for example.

This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the sender immediately by e-mail and delete this e-mail permanently.