Re: ndb OPTIMIZE - komt...@livetext.com - com.mysql.lists.cluster

6 messages in com.mysql.lists.clusterRe: ndb OPTIMIZE

From	Sent On	Attachments
samuel	08 Sep 2006 02:08
Mikael Ronström	08 Sep 2006 03:24
samuel	08 Sep 2006 03:57
Apostolos Pantsiopoulos	08 Sep 2006 13:07
komt...@livetext.com	08 Sep 2006 18:51
Apostolos Pantsiopoulos	09 Sep 2006 06:46

Subject:	Re: ndb OPTIMIZE
From:	komt...@livetext.com (komt...@livetext.com)
Date:	09/08/2006 06:51:51 PM
List:	com.mysql.lists.cluster

In my testing I needed to drop the table before the memory was reclaimed, I suspect that ndbd would release unused memory at some appropriate point like gc of java.

-kem

On Fri, September 8, 2006 3:07 pm, Apostolos Pantsiopoulos wrote:

I also made some tests a few days ago just to identify if such a problem existed. The conclusion that I drew was that ndb freed all the used memory after deleting all the rows in a table. I don't know if there are any memory leaks that could bring up memory-full issues after a lot of such transactions, but in a simple test environment I could not find any. This however is a very important issue and I would find it very helpful if someone from MySQL AB gave us an answer. MySQL NDB Cluster using a lot of memory is one thing, but not being able to identify your RAM needs is another. It could lead to unexpected (and disastrous) behaviour.

samuel wrote:

2006/9/8, Mikael Ronström <mik...@mysql.com>:

Hi, In 4.1 and 5.0 you can optimize by performing a rolling restart of the nodes, meaning restart one ndbd at a time.

Rgrds Mikael

2006-09-08 kl. 11.09 skrev samuel:

Hi all,

I'm just starting with myqsl cluster so first of all apologies if the question have already been answered. I've tried to find the answer but I couldn't...any link would be higly apreciated.

I'm using a compiled version of mysql 5.0.18 and I was wondering how to "emulate" the OPTIMIZE-like functions in the ndb storage engine. I recently deleted a large number of entries and I don't know if there's something I have to manually execute or the ndb cluster takes care itself of "everything". The "problem" is that even thoguh I have deleted the entries, I have run out of index memory and I can not (or it seems) to reuse this memory: my mysql APIs gives random errors about table full (sometimes it works and sometimes it does not).

I'm on the process of extending the memory but also considering the stability/performance of the disk storage for mysql 5.1 NDB...any link where I can get more info about this???

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

-------------------------------