|
More site info...
|
|
Forum profile page for Cluster on http://www.mysql.com.
This report page is the aggregated overview from a single forum: Cluster, located on the Message Board at http://www.mysql.com.
This forum profile page summarizes the general forum statistics such as: Users Activity, Forum Activity, and Top Authors, which are reported in either a table or graph below for a given reporting time period.
Additional forum profile information for "Cluster" on the Message Board at http://www.mysql.com is also shown in the following ways:
1) Latest Active Threads
2) Hot Threads for Last Week
Warning: These statistics are generated using 'best efforts' and can experience delays and reporting errors at times. Please note that such statistics do not constitute a forum's popularity and/or exact posting volumes at any given reporting period.
|
|
|
|
|
Posting activity on Cluster:
|
|
Week
|
Month
|
3 Months
|
|
Threads:
|
10
|
60
|
254
|
|
Post:
|
13
|
103
|
420
|
|
|
Cluster Posting activity graph:
|
Top authors during last week:
user's latest post:
Re: no database replication on...
Published (2009-11-26 18:15:00)
Hello Dan, Can you try the same setup with NoOfReplicas=2 instead? Using 4 replicas is not well tested should be avoided for production. You may be running into a bug where not all the data nodes are getting all the right information distributed to them. The SQL nodes then would not see all the correct data if the data nodes do not all contain all tables. If the table appear correctly on all SQL nodes when Using NoOfReplicas=2 submit a bug...
user's latest post:
Re: this is very large table
Published (2009-11-23 07:24:00)
Hello Kim, It looks like you have hit the max rows pointer. Please try the following: CREATE TABLE `TEST` ( `A` date NOT NULL, `B` varchar(255) NOT NULL, `C` varchar(50) NOT NULL, `D` varchar(50) NOT NULL, `E` varchar(50) NOT NULL, KEY `B` (c_autoquery_nosp (10)) ) TABLESPACE ts_1 STORAGE DISK ENGINE=ndbcluster MAX_ROWS=10000000000 DEFAULT CHARSET=utf8; -- Andrew Hutchings, MySQL Support Engineer, Americas Sun Microsystems, United Kingdom...
user's latest post:
ndb and max_rows
Published (2009-11-21 08:50:00)
Hello everyone, I've got a problem with partitions in a mysql cluster (7.0.8a under centos 5.3) with 4 data nodes. I have a big table which I want to create with 4 partitions. The problem is I've got the following error in one of the data nodes while inserting rows: Status: Temporary error, restart node Message: Array index out of range (Internal error, programming error or missing error message, please report a bug) Error: 2304...
Francesco Domenico riggi
1
user's latest post:
How check how many row get a lock?
Published (2009-11-23 04:32:00)
Hello I am wondering which way is possible in mysql cluster to check each table status and in particular if on the same table some row get locked and how many times. I found ndb_show_tables command bu I am not sure about "slock" in ndb_schema.. is that what I am looking for? Thanks Franco
user's latest post:
Re: Recommended node layout
Published (2009-11-24 00:14:00)
Hi Andrew Thank you for those suggestions - we are implementing and testing currently Another question, though. Are the mysqld processes memory intensive on the SQL nodes? I am thinking of moving some RAM between the servers (Instead of 48/48/48/48, I am thinking of 24/24/72/72, with 24 for the mysqld nodes and 72 for the ndb nodes) Would having equal ram in the SQL/NDB nodes help, or can I dedicate more to ndb storage?
user's latest post:
MySQL Cluster downgrade
Published (2009-11-27 05:16:00)
Hi, I have a system with MySQL Cluster 6.3.27a, I plan to upgrade the system to the latest MySQL Cluster 7.x version, I have some questions about the upgrade and downgrade According to the documentation the online upgrade from version 6.3.27a to 7.x version is supported, but I haven´t found any information about the downgrade from 7.x release back to 6.3 release. Is this downgrade possible? Regarding the backup and restore, Can I restore a...
user's latest post:
Re: no database replication on...
Published (2009-11-27 02:25:00)
Just wanted to point out that one of the reasons we don't do much testing with NoOfReplicas > 2 is that 2 replicas are sufficient for providing HA. Jon Stephens Technical Writer / MySQL Documentation Team Sun Microsystems / MySQL and Software Infrastructure Group Stockholm, Sweden MySQL Sun
user's latest post:
Re: too many partitions??
Published (2009-11-26 09:05:00)
Thank you for the information. These 8 partitions still worry me. According to "explain partitions" when i had 2 data nodes, the selects of the primary keys went directly to the adecuate partition. With 4 data nodes and these strange 8 partitions the "explain partitions" says that the same select involves all partitions.
|
|
|
|
Latest active threads on Cluster::
Started 5 days, 22 hours ago (2009-11-26 18:15:00)
by Matthew Montgomery
Hello Dan,
Can you try the same setup with NoOf replicas=2 instead? Using 4 replicas is not well tested should be avoided for production. You may be running into a bug where not all the data nodes are getting all the right information distributed to them. The SQL nodes then would not see all the correct data if the data nodes do not all contain all tables.
If the table appear ...
Started 5 days, 22 hours ago (2009-11-26 18:01:00)
by Matthew Montgomery
When you switched from 2 node to 4 nodes, did you also change from ndbd to ndbmtd? If you were already running ndbmtd with 2 nodes did you increase NoOfExecutionThreads when you added the other 2 ndbmtd nodes? NoOfExecutionThreads=4 will cause 2 partitions to be created per node.
http://dev.mysql.com/doc/ndbapi/en/ndb-internals-n dbd-filesystemdir-files.html
Edited 1 time(s). Last ...
Started 1 month ago (2009-10-31 08:44:00)
by Matthew Montgomery
I'd be interested to see what the ndb_3_error.log and ndb_4_error.log look like along with any trace files the nodes generated when they crashed. This will tell us the real cause of the crash and weather it has been fixed in a later version. (7.0.8a)
Started 2 weeks, 5 days ago (2009-11-13 12:57:00)
by Andrew Morgan
I'd suggest starting with 1 multi-threaded data node on each of 2 servers (use ndbmtd and set MaxNoOfExecutionThreads to 8).
Run a management node together with at least 1 MySQL Server on each of the other 2 machines. For the mysqld processes, try using the ndb-cluster-connection-pool parameter to have mulitple connections to the data nodes for greater through put if you have a multi-...
Started 1 week, 2 days ago (2009-11-23 10:41:00)
by Matthew Montgomery
Currently there isn't a way to expose individual row locks. The "slock" indicates weather or not that particular schema operation has been successfully written to the binlogs. When an CREATE/DROP/ALTER are issued, the cluster the SQL statement and some other information about it into ndb_schema. Along with this information cluser writes a bitmap of all nodes listening to the binlog subscriber ...
Started 1 week, 2 days ago (2009-11-23 07:24:00)
by Andrew Hutchings
Hello Kim,
It looks like you have hit the max rows pointer. Please try the following:
CREATE TABLE `TEST` (
`A` date NOT NULL,
`B` varchar(255) NOT NULL,
`C` varchar(50) NOT NULL,
`D` varchar(50) NOT NULL,
`E` varchar(50) NOT NULL,
KEY `B` (c_autoquery_nosp (10))
) TABLESPACE ts_1 STORAGE DISK ENGINE=ndbcluster MAX_ROWS=10000000000 DEFAULT CHARSET=utf8;
--
Andrew ...
Started 1 week, 5 days ago (2009-11-20 10:40:00)
by Andrew Hutchings
Hello Ben,
Can you please specify which version you are using?
--
Andrew Hutchings, MySQL Support Engineer, Americas
Sun Microsystems, United Kingdom
http://www.sun.com/mysql/
Started 1 week, 6 days ago (2009-11-19 08:57:00)
by Andrew Hutchings
Show Warnings needs to be done in the mysql node, not the data node. It needs to be executed straight after the error occurs and before any other statements.
As a bare minimum you should get the same error repeated in the response. In cluster this error code can happen for one of two reasons:
1. IndexMemory is full
2. MaxNoOfAttributes is full
You can test for number one with ...
|
|
Hot threads for last week on Cluster::
Started 5 days, 22 hours ago (2009-11-26 18:15:00)
by Matthew Montgomery
Hello Dan,
Can you try the same setup with NoOf replicas=2 instead? Using 4 replicas is not well tested should be avoided for production. You may be running into a bug where not all the data nodes are getting all the right information distributed to them. The SQL nodes then would not see all the correct data if the data nodes do not all contain all tables.
If the table appear ...
Started 5 days, 22 hours ago (2009-11-26 18:01:00)
by Matthew Montgomery
When you switched from 2 node to 4 nodes, did you also change from ndbd to ndbmtd? If you were already running ndbmtd with 2 nodes did you increase NoOfExecutionThreads when you added the other 2 ndbmtd nodes? NoOfExecutionThreads=4 will cause 2 partitions to be created per node.
http://dev.mysql.com/doc/ndbapi/en/ndb-internals-n dbd-filesystemdir-files.html
Edited 1 time(s). Last ...
Started 2 weeks, 5 days ago (2009-11-13 12:57:00)
by Andrew Morgan
I'd suggest starting with 1 multi-threaded data node on each of 2 servers (use ndbmtd and set MaxNoOfExecutionThreads to 8).
Run a management node together with at least 1 MySQL Server on each of the other 2 machines. For the mysqld processes, try using the ndb-cluster- connection-pool parameter to have mulitple connections to the data nodes for greater through put if you have a multi-...
Started 1 month ago (2009-10-31 08:44:00)
by Matthew Montgomery
I'd be interested to see what the ndb_3_error.log and ndb_4_error.log look like along with any trace files the nodes generated when they crashed. This will tell us the real cause of the crash and weather it has been fixed in a later version. (7.0.8a)
|
|
Size: 526 bytes
