Re: nodes on two seperate lans hangs

Hi Vincent.

> My cluster, with two data nodes, works fine when all nodes are on the
> same network. But when configuring the cluster exactly the same way but
> replacing a data node with a node on a remote network, it hang up in
> starting phase (I did the --initial start after modifying the conf).
> Maybe you can help or give me some hints

If I were you, I would use lsof and tcpdump/ngrep to find out exactly on what interfaces/IP addresses and ports the ndbd processes are listening and to what addresses and ports they try to connect, I would also see if I can connect to those IP addresses and ports using telnet.

> manager is on host NET1.5
>
> -- NDB Cluster -- Management Client --
> ndb_mgm> show ;
> Connected to Management Server at: localhost:1186
> Cluster Configuration
> ---------------------
> [ndbd(NDB)] 2 node(s)
> id=3 @NET1.6 (Version: 5.1.20, starting, Nodegroup: 0)
> id=7 @NET2.178 (Version: 5.1.20, starting, Nodegroup: 0)
>
> [ndb_mgmd(MGM)] 1 node(s)
> id=1 (Version: 5.1.20)
>
> [mysqld(API)] 3 node(s)
> id=4 (not connected, accepting connect from NET1.6)
> id=6 (not connected, accepting connect from NET1.7)
> id=8 (not connected, accepting connect from NET2.178)
>
>
> Remark: I am suprised here I have no master ? When all machines are on
> the same lan NET1.0/24, node 3 becomes master and API nodes connect.

I think because the data nodes can not contact each other. If you fix the problem with communication then they will elect the master.

> I get messages in the ndb_mgmd log I don't understand:
>
> 2007-07-25 14:09:30 [MgmSrvr] INFO -- Node 3: Initial start, waiting
> for 0000000000000080 to connect, nodes [ all: 0000000000000088
> connected: 0000000000000008 no-wait: 0000000000000000 ]
> 2007-07-25 14:09:32 [MgmSrvr] INFO -- Node 7: Initial start, waiting
> for 0000000000000008 to connect, nodes [ all: 0000000000000088
> connected: 0000000000000080 no-wait: 0000000000000000 ]
>
> How do I interpret "0000000000000080" and "0000000000000008" as machine
> identifiers ? Can you give details on these identifiers ?

It is something like this. They are not machine identifiers, they are nodes and node sets. Node 1 is encoded as 0000000000000001, Node 2 is encoded as 0000000000000002, Node 3 is encoded as 0000000000000004, Node 4 is encoded as 0000000000000008, Node 5 - 0000000000000010 and so on.

> NB: firewall is wide open between all these hosts (nothing is blocked,
> tcpdump report a lot of traffic between manager and nodes, but no
> traffic at all between the 2 data nodes).
> NB: If I revert to a "all nodes on the same lan" config, evrything
> restart normally.
> NB: I checked the configuration for possible stupid errors, but this
> looks nice right now.
> NB: hosts on seperate lans have 100Mb/s connections between them.
>
> My final question is: "Even if not recommended, is it possible to have
> such a configuration with data nodes on seprate networks ?"

Do you need help?

In addition to our free technical library, Pantek provides professional IT services to companies who utilize Open Source software: Learn more about Pantek's Technical Support offerings Place an order for 24/7/365 Technical Support here Learn more about Pantek, who we are, and what we do Chat with a Live Operator Related lists.mysql.com Links announce benchmarks bugs certification cluster commits dotnet gui-tools internals java maxdb myodbc mysql mysqldoc packagers perl plusplus replication win32 Request more information about Pantek

If the connection between the networks is really 100 Mb/s then it should be possible.

Regards,

Anatoly.

-- 
MySQL Cluster Mailing List
For list archives: 
http://lists.mysql.com/cluster
To unsubscribe:    
http://lists.mysql.com/cluster?unsub=lists@pantek.com