5 messages in com.mysql.lists.clusterRe: multiple management nodes?| From | Sent On | Attachments |
|---|---|---|
| Jaime Peñalba | 21 Oct 2004 08:08 | |
| Steve Reaser | 21 Oct 2004 08:13 | |
| Tomas Ulin | 21 Oct 2004 15:32 | |
| Jaime Peñalba | 27 Oct 2004 05:41 | |
| Jaime Peñalba | 27 Oct 2004 05:44 |
| Subject: | Re: multiple management nodes? |
|---|---|
| From: | Jaime Peñalba (jpen...@gmail.com) |
| Date: | 10/27/2004 05:44:30 AM |
| List: | com.mysql.lists.cluster |
Sorry for sending this mail several times, but it was incomplete.
Im trying to configure two management nodes but it cannot get it working, and i dont know what is wrong.
This is what im doing: - start ndb_mgmd on both servers "../../libexec/ndb_mgmd -c config.ini" - start ndbd on both servers ../../libexec/ndbd -c "host=apache1-db:2200,apache2-db:2200;nodeid=3" ../../libexec/ndbd -c "host=apache1-db:2200,apache2-db:2200;nodeid=4"
But ndbd daemons are conecting to localhost mgmd, i have tried also to conect ndbd from apache2-db to apache1-db mgm server using "../../libexec/ndbd -c "host=apache1-db:2200"" but it still conects to localhost mgmd
apache1-db# echo "show" | ndb_mgm -- NDB Cluster -- Management Client -- Connecting to Management Server: localhost:2200 Cluster Configuration
--------------------- [ndbd(NDB)] 2 node(s) id=3 @127.0.0.1 (Version: 3.5.3, starting, Nodegroup: 0) id=4 (not connected, accepting connect from apache2-db.ccoo.es)
[ndb_mgmd(MGM)] 2 node(s) id=1 @192.168.0.2 (Version: 3.5.3) id=2Unable to find node with id: 2 [mysqld(API)] 2 node(s) id=5Unable to find node with id: 5
apache2-db# echo "show" | ndb_mgm -- NDB Cluster -- Management Client -- Connecting to Management Server: localhost:2200 Cluster Configuration
--------------------- [ndbd(NDB)] 2 node(s) id=3 (not connected, accepting connect from apache1-db.ccoo.es) id=4 (Version: 3.5.3, starting, Nodegroup: 0)
[ndb_mgmd(MGM)] 2 node(s) id=1Unable to find node with id: 1 [mysqld(API)] 2 node(s) id=5Unable to find node with id: 5
Here is my config.ini for both ndb_mgmd
[COMPUTER] ID: 1 HostName: apache1-db
[COMPUTER] ID: 2 HostName: apache2-db
[NDBD DEFAULT] NoOfReplicas: 2
[MYSQLD DEFAULT] [NDB_MGMD DEFAULT] DataDir: /usr/local/mysql/var/mysql-cluster
[TCP DEFAULT]
[NDB_MGMD] Id: 1 ExecuteOnComputer: 1
[NDB_MGMD] Id: 2 ExecuteOnComputer: 2
[NDBD] Id: 3 ExecuteOnComputer: 1 DataDir: /usr/local/mysql/var/mysql-db
[NDBD] Id: 4 ExecuteOnComputer: 2 DataDir: /usr/local/mysql/var/mysql-db
[MYSQLD] Id: 5 ExecuteOnComputer: 1
[MYSQLD] Id: 6 ExecuteOnComputer: 2
I dont know what im doing wrong...
Thanks, Jaime.
On Thu, 21 Oct 2004 22:33:10 +0000, Tomas Ulin <tom...@mysql.com> wrote:
Steve Reaser wrote:
The other question is about HA, as the cluster configuration is two node only, if the node with the management server fails, the other node will still work but it couldnt be restarted. is this right?
I've been told that nodes can be restarted without the management server. They store their config locally upon first startup.
It is possible to run it almost this way, but it is not the default behaviour, and it is not the recommended way.
The management server is default needed for arbitration and should run on a different computer from the ndbd nodes.
A mysqld can also function as an arbitrator, in which case it is not so sensitive where the management server is running. see config parameter ArbitrationRank in documentation
They only need the management server to all get up and running initially.
You normally need it to restart nodes. If you want a different behaviour it's outsite current documentation.
Else there is any way to have more than one management node? and how i can tell the ndbd daemon to use both management nodes? If this isnt possible the cluster has only 1 failure point to stop working. There is any solution for this?
I don't think multiple management nodes is possible, but since it's only needed for initial startup and could be monitored easily it's not really a single point of failure as the rest of the system will work including nodes dropping in and out without it.
multiple mgmt nodes is possible yes. Some gotha's are there though:
1. You need to make sure they all have the same configuration 2. You must specify nodeid's specifically for them and the other nodes and that you don't use the same id's twice
e.g. mysqld --ndbcluster --ndb-connectstring="nodeid=12,ndb_mgmd_host1:2200,,ndb_mgmd_host2:2200,ndb_mgmd_host3:2200"
They will be contacted in order...
-Steve