Re: Cluster statistics - Mikael Ronström - com.mysql.lists.cluster

6 messages in com.mysql.lists.clusterRe: Cluster statistics

From	Sent On	Attachments
Christian Schramm	14 Sep 2006 01:30
Ole Vedel Villumsen	14 Sep 2006 01:45
Christian Schramm	14 Sep 2006 02:12
Mikael Ronström	14 Sep 2006 07:57
Adam Dixon	14 Sep 2006 15:59
Mikael Ronström	14 Sep 2006 23:27

Subject:	Re: Cluster statistics
From:	Mikael Ronström (mik...@mysql.com)
Date:	09/14/2006 11:27:22 PM
List:	com.mysql.lists.cluster

Hi,

2006-09-15 kl. 00.59 skrev Adam Dixon:

This actually could happen if you have restarted one node, but not the other, and the remaining long standing node still thinks that it is using all that data. Depending on how much your data set changes. I delete around 3-400,000 rows every night so If I restart 1 node but not the others, the freshly restarted one may say 66% but the long standing nodes may still say 79% as an example.

Quite agree, didn't think of that.

Rgrds Mikael

On 9/15/06, Mikael Ronström <mik...@mysql.com> wrote:

Hi,

2006-09-14 kl. 11.13 skrev Christian Schramm:

"all dump 1000" did it. Thanks a lot. I noticed a big difference between the data usage (46% on node1, 66% on node2). Is this normal? The index usage on both nodes is almost the same (around 30%).

No, doesn't sound very normal unless DataMemory on the two nodes is different. There's been one bug fix in this are with regards to BLOB's that put all data in the first partition but I presume in your case node1 and node2 are handling the same data.

Rgrds Mikael

Ole Vedel Villumsen schrieb:

In the management console, issue the command ALL DUMP 1000. This should tell you something like "Sending dump signal with data: 0x00zzzzzz" a couple of times. It leaves some lines in the ndb log file - use tail -4 ndb_1_cluster.log. As far as I've understood, the numbers after "of total" give the memory usage for data and indexes, respectively, measured in kilobytes. The number in the ALL DUMP 1000 command has to be 1000, or strange results may occur, I've read.

Someone else should be able to tell you about the nbd_size.pl script if you don't already know it.

Hope it helps.

Yours, Ole

-----Original Message----- From: Christian Schramm [mailto:mail...@dretschler.com] Sent: 14. september 2006 10:31 To: clus...@lists.mysql.com Subject: Cluster statistics

Hi,

i have a short question. I'm managing a cluster consisting of 2 APIs and 2 nodes. The nodes each have 6GB of RAM. There are currently running 2 databases (both together about 120 tables).

So, i'd like to add another database to the cluster, wich will first not contain such a data amount as it is in the other databases. The new database will contain 18 tables.

In the cluster configuration there is a certain memory reserved for data, indexes etc. I can see the load and the memory usage of the whole server, but I don't see how much of the data memory is currently used.

Is there a possibility to read this out somewhere or to create statistics? The API logs only give me alerts, and information about processed backups, checkoints etc.

Kind regards C. Schramm

Mikael Ronstrom, Senior Software Architect MySQL AB, www.mysql.com

Jumpstart your cluster: http://www.mysql.com/consulting/packaged/cluster.html My blog: http://mikaelronstrom.blogspot.com

Mikael Ronstrom, Senior Software Architect MySQL AB, www.mysql.com

Jumpstart your cluster: http://www.mysql.com/consulting/packaged/cluster.html My blog: http://mikaelronstrom.blogspot.com