| Bug #28292 | ndbcluster fails to initialize when my.cnf defines ndb-nodeid | ||
|---|---|---|---|
| Submitted: | 8 May 2007 0:30 | Modified: | 22 Mar 2008 10:55 |
| Reporter: | Gerry Reno | Email Updates: | |
| Status: | Not a Bug | Impact on me: | |
| Category: | MySQL Cluster: Cluster (NDB) storage engine | Severity: | S3 (Non-critical) |
| Version: | 5.1.17-beta, 5.1.22, 5.1.24-BK | OS: | Any |
| Assigned to: | CPU Architecture: | Any | |
[8 May 2007 0:30]
Gerry Reno
[8 May 2007 15:26]
Hartmut Holzgraefe
can't reproduce, works fine for me ... are you sure that node id 5 is configured as a [mysqld] slot in the cluster configuration and that it is not already allocated by some other mysqld process or configured for a different host? could you provide your config.ini file and the hostname/ip of the failing mysqld node?
[8 May 2007 15:43]
Gerry Reno
Here is mgm output: ndb_mgm> show Cluster Configuration --------------------- [ndbd(NDB)] 2 node(s) id=2 @192.168.1.40 (Version: 5.1.17, Nodegroup: 0) id=3 @192.168.1.41 (Version: 5.1.17, Nodegroup: 0, Master) [ndb_mgmd(MGM)] 1 node(s) id=1 @192.168.1.40 (Version: 5.1.17) [mysqld(API)] 2 node(s) id=4 (not connected, accepting connect from any host) id=5 @192.168.1.25 (Version: 5.1.17) As you can see nodeid 5 is a mysqld (shown is the output when I start the server without specifying the ndb-nodeid). Here is config.ini: [TCP DEFAULT] portnumber=2202 #-- NDB_MGMD MANAGEMENT DAEMON: [NDB_MGMD] #hostname=192.168.1.25 # Hostname or IP address of MGM node #datadir=/opt/lampp/var/lib/mysql-cluster # Directory for MGM node log files hostname=192.168.1.40 # Hostname or IP address of MGM node datadir=/var/lib/mysql-cluster # Directory for MGM node log files #-- NDBD DATA NODE DEFAULTS: [NDBD DEFAULT] NoOfReplicas=2 # Number of replicas # GR: initial allocation: 600M out of 2GB total for MySQL data/indexes DataMemory=500M # How much memory to allocate for data storage (default=80M) IndexMemory=100M # How much memory to allocate for index storage (default=18M) #-- NDBD DATA NODES: # one [NDBD] section per data node [NDBD] hostname=192.168.1.40 # Hostname or IP address datadir=/usr/local/mysql/data # Directory for this data node's data files [NDBD] hostname=192.168.1.41 # Hostname or IP address datadir=/usr/local/mysql/data # Directory for this data node's data files #-- SQL (API) NODES: [MYSQLD] [MYSQLD] hostname=192.168.1.25 The ip of the failing node is 192.168.1.25.
[8 May 2007 16:20]
Hartmut Holzgraefe
i've now changed my test configuration to not use explicit node IDs but i still can't reproduce your problem, the mysqld servers start up just fine for me could you add explicit ID= lines for all nodes, too, and see if it makes a difference in your setup?
[8 May 2007 16:38]
Gerry Reno
I changed my.cnf to the following: [MYSQLD] ndbcluster # run NDB storage engine ndb-connectstring=192.168.1.40:1186 # location of management server ndb-mgmd-host=192.168.1.40 ndb-nodeid=5 #this caused the ndbcluster engine to not be loaded by the plugin - see log engine_condition_pushdown=1 # send WHERE statements to Data Nodes for evaluation [NDBD] ndbd-host=192.168.1.40 ndb-nodeid=2 [NDBD] ndbd-host=192.168.1.41 ndb-nodeid=3 [NDB_MGMD] ndb-mgmd-host=192.168.1.40 ndb-nodeid=4 # Options for all cluster ndbd processes: [MYSQL_CLUSTER] ndb-connectstring=192.168.1.40:1186 The server still fails to start, producing the exact same error as previously listed.
[8 May 2007 18:48]
Gerry Reno
History of my installation: Original installation: 5.0.37 with ndbcluster. Upgrade installation: 5.1.17-beta with ndbcluster. Ran mysql_upgrade script immediately after upgrade install. Brought up mgmd and ndbd nodes. Attempted to start SQL Node and that is where the failed init for ndbcluster happened whenever a sql nodeid is specified in my.cnf. Hope this helps.
[28 Jan 2008 11:25]
Valeriy Kravchuk
Please, try to repeat with a newer version, 5.1.22, and inform about the results.
[6 Feb 2008 15:21]
Stanisław Pitucha
Same thing happens to me: [mysqld] ndbcluster ndb-connectstring=1.2.3.4;1.2.3.5 #ndb-nodeid=4 If I remove comment, I get message saying: Could not initialize handle to management server: Illegal connect string : nodeid=4,nodeid=4,1.2.3.4;1.2.3.5 Management node has only sections like: [MYSQLD] Hostname = 1.2.3.4 No node ids, only hostname for every entry. Configured with 2*data,2*mgm,2*api I'm running 5.1.22 rhel5 x64 packages from mysql site.
[11 Feb 2008 18:51]
Jeff Wang
I also have this problem using version 5.1.24 from BK. I compiled from source to run 64-bit on Mac 10.5. Same set up and config with id's defined in config.ini. Sometimes I can connect but I mostly I get an error, so the problem seems intermittent. When I get a successful connection, the data nodes acknowledge the mysql nodes right away. When it get an error, it seems as though the data nodes cannot communicate with the mysql node. This is seen in the mgmt cluster logs.
[29 Feb 2008 0:00]
Bugs System
No feedback was provided for this bug for over a month, so it is being suspended automatically. If you are able to provide the information that was originally requested, please do so and change the status of the bug back to "Open".
[22 Mar 2008 10:55]
Sveta Smirnova
Thank you for the report. I was able to repeat described behavior with following layout: $cat etc/ndb_mgmd.cfg ... [MYSQLD] NodeId = 9 [MYSQLD] $cat etc/ndb_node_api1.cfg [MYSQL_CLUSTER] ndb-connectstring=127.0.0.1:35118 # The MySQL server [mysqld] ndbcluster ndb-connectstring=127.0.0.1:35118 port=35117 socket=/tmp/mysql_ndb_api1.sock datadir=/Users/apple/Desktop/cluster_demo/data3 log-error $cat etc/ndb_node_api2.cfg [MYSQL_CLUSTER] ndb-connectstring=127.0.0.1:35118 # The MySQL server [mysqld] ndbcluster ndb-connectstring=127.0.0.1:35118 port=35119 socket=/tmp/mysql_ndb_api2.sock datadir=/Users/apple/Desktop/cluster_demo/data4 ndb-nodeid=10 log-error In this case API node without specified ndb-nodeid sometime gets nodeid = 10, so nodeid=10 can not be allocated to another node and leads to plugin 'ndbcluster' initialisation failure. In other cases API node without specified ndb-nodeid gets nodeid = 9 and second node starts successfully. So I close the report as "Not a Bug".
[23 Mar 2008 17:13]
Stanisław Pitucha
I'm not sure if that was the problem. Please reopen if not. This happens also when *all* nodes have explicit id set. (also with hostnames) My full config: --- config.ini --- DataDir = /var/lib/mysql-cluster NoOfReplicas = 2 DataMemory = 1024M IndexMemory = 100M [NDB_MGMD] Id = 1 Hostname = A.B.C.D [NDB_MGMD] Id = 2 Hostname = A.B.C.E [MYSQLD] Id = 3 Hostname = A.B.C.F [MYSQLD] Id = 4 Hostname = A.B.C.G [NDBD] Id = 5 Hostname = A.B.C.H [NDBD] Id = 6 Hostname = A.B.C.I --- --- --- my.cnf --- [mysqld] ndbcluster ndb-connectstring=A.B.C.D;A.B.C.E #ndb-nodeid=3 skip-innodb default-table-type=ndbcluster --- --- same on second one but with 'ndb-nodeid=4' Works with 'ndb-nodeid' commented out, but when I uncomment it (doesn't matter if config.ini has 'Hostname:' part commented out, or not) api node stops with error reported before (bad connect string). I believe this is different situation from what previous commenter tested.
[24 Mar 2008 21:57]
Sveta Smirnova
Stanislaw, thank you for the feedback. Do you meet the problem when ndb-nodeid set for both mysqld or only for one?
[25 Mar 2008 10:27]
Stanisław Pitucha
Actually I've just checked that it doesn't depend on server / connection at all. I've tried to start api node (mysqld --user=mysql) with configuration like before (exactly those addresses - I've tried using 1.2.3.4;1.2.3.5): --->8--- [mysqld] ndbcluster ndb-connectstring=1.2.3.4;1.2.3.5 ndb-nodeid=3 skip-innodb default-table-type=ndbcluster --->8--- Result: --->8--- 080325 10:19:59 [Note] Plugin 'InnoDB' disabled by command line option Could not initialize handle to management server: Illegal connect string : nodeid=3,nodeid=3,1.2.3.4;1.2.3.5 080325 10:19:59 [ERROR] Plugin 'ndbcluster' init function returned error. 080325 10:19:59 [ERROR] Plugin 'ndbcluster' registration as a STORAGE ENGINE failed. 080325 10:19:59 [ERROR] Unknown/unsupported table type: ndbcluster 080325 10:19:59 [ERROR] Aborting 080325 10:19:59 [Note] /usr/sbin/mysqld: Shutdown complete --->8--- It has failed before it tried to send any mysql-cluster packet at all - there was no ndb communication (monitored tcpdump). Exactly the same result as with previous tries in 'real' cluster.
