7 messages in com.mysql.lists.clusterRe: Force master| From | Sent On | Attachments |
|---|---|---|
| Andrew Hutchings | 15 Feb 2007 06:32 | |
| Geert Vanderkelen | 15 Feb 2007 06:50 | |
| Andrew Hutchings | 15 Feb 2007 06:57 | |
| Geert Vanderkelen | 15 Feb 2007 07:07 | |
| Hartmut Holzgraefe | 15 Feb 2007 07:12 | |
| Andrew Hutchings | 15 Feb 2007 07:29 | |
| Matthew Montgomery | 15 Feb 2007 21:28 |
| Subject: | Re: Force master![]() |
|---|---|
| From: | Matthew Montgomery (mmon...@mysql.com) |
| Date: | 02/15/2007 09:28:38 PM |
| List: | com.mysql.lists.cluster |
On Thu, 2007-02-15 at 15:30 +0000, Andrew Hutchings wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Geert Vanderkelen wrote:
No, the data nodes decide between themselves who continues and who not. Important in case of network failure is arbitration which is done by the management node (or the sql nodes), on another physical machine (preventing split brains). If the master fails, a new one will be elected.
A bit weird that you want a node to go down first in case of problems.. It should not matter as they should be completely equal.. in a nice Cluster environment.
Ah, OK, makes sense. Situation is as follows:
All but 1 data and sql node (on the same server) are on the same switch, the link between the 2 switches had a large packet loss a few days ago and the node on the separate switch failed one of the other nodes before it was failed by the management node.
Obviously the cluster can't know the situation it is in physically so I was more wondering if you could force what I thought was the master, but is actually the arbitrator on to one of the nodes on the populated switch.
Yes, in the config.ini you can specify the ArbitrationRank = 0 for any node you want never to act as an arbitrator. A value of 1 makes it have high priority as an arbitrator and 2 gives it low priority.
So all the API/Management nodes on the switch with the most DATA nodes on it should have ArbitrationRank = 1 while the nodes on the second switch get ArbitrationRank = 2 or 0
In the end all the nodes failed due to: http://bugs.mysql.com/bug.php?id=24447 but that is another story.
Regards Andrew - -- Andrew Hutchings - Linux Jedi - http://www.linuxjedi.co.uk/ A-Wing Internet Services - http://www.a-wing.co.uk/ Windows is the path to the darkside...Windows leads to Blue Screen. Blue Screen leads to downtime. Downtime leads to suffering...I sense much Windows in you. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.5 (Darwin) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
iD8DBQFF1HyGZ+2Y3WLvbJIRAkvEAKCPF/x6XU/HDOkha8+DV2DkJfs76wCdGnvd yb7h/QvMcC8gQqqPF0AMwUM= =oJ3Z -----END PGP SIGNATURE-----
-- Matthew Montgomery, Support Engineer MySQL Inc, www.mysql.com
Are you MySQL certified? www.mysql.com/certification




