There must be an even number of replicas because the cluster mirrors data between two machines. It doesn't do "three way" mirroring. An even number of nodes are required because each two data nodes becomes a node group.
If you have three machines, you could create two NDB processes on each machine each each using some of the memory you want to dedicate to NDB, though this adds some additional overhead in the allocation of operation records and what not. Just make sure you don't mirror data to the same machine. I (think) they are mirrored in the order they occur so do something like the following (collapsed from standard INI format for brevity) datamemory 33M, indexmemory 17M MGM id 1, machine 1 DB id 2, machine 1 DB id 3, machine 2 DB id 4, machine 3 DB id 5, machine 1 DB id 6, machine 2 DB id 7, machine 3 API id 8, machine 1 API id 9, machine 2 API id 10, machine 3 I haven't actually tested that configuration, so your milage may vary. If you do test it, then let me know how well it works for you. On Mon, 9 Aug 2004 12:07:11 -0400, Wheeler, Alex <[EMAIL PROTECTED]> wrote: > After doing further testing, it would appear that the number of nodes must be a > power of 2: > A test of 2 nodes works fine, but 3 fails > 4 works, but 6 fails > 8 works... > Is this documented and I just missed it? > > -- > Alex Wheeler > > > > > -----Original Message----- > From: Mikael Ronström [mailto:[EMAIL PROTECTED] > Sent: Thursday, August 05, 2004 1:15 PM > To: Wheeler, Alex > Cc: [EMAIL PROTECTED] > Subject: Re: Nodes crash on table creation > > Hi Alex, > I tried a similar configuration with all nodes on the same computer and > that worked fine. > Please provide the trace file in some manner and I'll check it once > more. > The other nodes fail since there is only one replica and thus when one > node fails then > the cluster fails. > > Rgrds Mikael > > 2004-08-05 kl. 18.34 skrev Alex Wheeler: > > > I am using the mysql-4.1.4-beta-nightly-20040804 snapshot, compiled > > with > > shared memory support, though I'm not using that feature yet. > > > > I'm using a 3 data node, 4 computer configuration, though I've tried > > several options and the only one that seems to work with this build is > > just 1 computer, here's my config file: > > > > [COMPUTER] > > Id:1 > > HostName: 192.168.208.101 > > > > [COMPUTER] > > Id:2 > > HostName: 192.168.208.100 > > > > [COMPUTER] > > Id:3 > > HostName: 192.168.208.102 > > > > [COMPUTER] > > Id:4 > > HostName: 192.168.208.103 > > > > [MGM] > > Id:1 > > ExecuteOnComputer: 1 > > ArbitrationRank: 1 > > > > [DB DEFAULT] > > NoOfReplicas: 1 > > IndexMemory: 50M > > DataMemory: 100M > > LockPagesInMainMemory: Y > > > > [DB] > > Id:6 > > ExecuteOnComputer: 2 > > FileSystemPath: /home/awheeler/cluster3/ndb_data6/ > > > > [DB] > > Id:7 > > ExecuteOnComputer: 3 > > FileSystemPath: /home/awheeler/cluster3/ndb_data7/ > > > > [DB] > > Id:9 > > ExecuteOnComputer: 4 > > FileSystemPath: /home/awheeler/cluster3/ndb_data9/ > > > > [API] > > Id: 20 > > ExecuteOnComputer: 1 > > ArbitrationRank: 2 > > > > > > > > The nodes come up fine, but when I try to create a table, at least one > > of the nodes will bail out and leave an error log: > > Current byte-offset of file-pointer is: 468 > > > > > > Date/Time: Thursday 5 August 2004 - 03:51:33 > > Type of error: error > > Message: Internal program error (failed ndbrequire) > > Fault ID: 2341 > > Problem data: DbtuxMeta.cpp > > Object of reference: DBTUX (Line: 128) 0x00000006 > > ProgramName: NDB Kernel > > ProcessID: 3244 > > TraceFile: ndb_7_trace.log.1 > > ***EOM*** > > > > The others exit with "Arbitrator decided to shutdown this node" > > > > The computers all have different amounts of memory, the smallest being > > 512MB. > > > > Let me know if you need more information, and where to send the trace > > file. > > > > Any ideas on what is causing ndbrequire to fail? > > > > -- > > MySQL Cluster Mailing List > > For list archives: http://lists.mysql.com/cluster > > To unsubscribe: > > http://lists.mysql.com/[EMAIL PROTECTED] > > > > > Mikael Ronström, Senior Software Architect > MySQL AB, www.mysql.com > > Clustering: > http://www.infoworld.com/article/04/04/14/HNmysqlcluster_1.html > > http://www.eweek.com/article2/0,1759,1567546,00.asp > > -- > MySQL Cluster Mailing List > For list archives: http://lists.mysql.com/cluster > To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED] > > -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]
