8 messages in com.mysql.lists.clusterRe: Memory issues with MySQL Cluster| From | Sent On | Attachments |
|---|---|---|
| Gbenga Bello | 12 Mar 2005 15:24 | |
| Mikael Ronström | 14 Mar 2005 01:11 | |
| Martin Skold | 14 Mar 2005 01:13 | |
| Ira Abramov | 14 Mar 2005 03:29 | |
| Alex Davies | 14 Mar 2005 03:35 | |
| Mikael Ronström | 14 Mar 2005 04:37 | |
| Martin Skold | 14 Mar 2005 04:45 | |
| Sancho2k.net Lists | 14 Mar 2005 06:25 |
| Subject: | Re: Memory issues with MySQL Cluster |
|---|---|
| From: | Mikael Ronström (mik...@mysql.com) |
| Date: | 03/14/2005 01:11:09 AM |
| List: | com.mysql.lists.cluster |
Hi,
2005-03-13 kl. 00.24 skrev Gbenga Bello:
Hi stewart and lists,
Thanks for the enlightenment of the other day. I have a few issues that I need to clarify on the memory usauge of MySQL Cluster and some other interface things. Here are the issues;
1)Having read the documentations on mysql cluster db, I found that mysql cluster by architecture is an in-memory database. Now does this means that it keeps all its data in memory (RAM). If this is so, at what point does it do this. Is it when the cluster is started that all data that were previously logged into disk is loaded into memory or could it be that it is as the data is requested for an application or client api. I would need detailed explanation of this part.
At start-up all data is loaded into memory.
2)If the cluster table data is stored in memory, does this mean I must have an equivalent amount of memory resource to handle this data. Take for instance, if I need to store like 20GB size of data, does this mean I must have 20GB total RAM capacity in mysql cluster to store this data?. Please I will need you to shed light on this.
Yes, all data must be in memory, so a 20 GB size of data means that you need 20 GB + overhead + possible replication into the ndbd nodes in the cluster.
3)How can I monitor the growth of my data in the cluster to determine when it becomes necesssary to increase memory resource. (RAM and Hard disk)
There is a event report when reaching 75%, 80% and so forth full. Check the MySQL Reference Manual for details.
4)If I run short or out of memory space what are the effects or danger this can pose on the performance of the cluster. Can this lead to a crash of the NDBD nodes or the entire cluster?
If the ndbd server starts swapping due to no memory then this can easily lead to crashes since ndbd processes use heartbeats to watch each other.
INTERFACE
===========
5)From my pilot installation of an 8 node mysql cluster(1 MGMD, 4NDBD, 3 MYSQLD), I observed that databases and tables created through one MySQLD cannot be view or related with through another MySQLD though they be of same member of the cluster
Databases are not "auto-discovered". Thus if you create a database you need to create it on all MYSQLD's. Tables are "auto-discovered" so they show up immediately after being created on another machine when requested by an SQL command.
6)How can I manage database and tables created from one interface (MySQLQ) using a different interface or how do I make several mysql prompt or mysqld view all database and tables residing on the cluster
Ensure that all databases created are created at all MYSQLD's. After that all MYSQLD's see the same view of the system.
7)How can I change the default Engine of MySQLD from MyISAM to NDBCLUSTER. Though I already have MySQLD configured for NDB Engine and this works.
There is a start-up parameter to MYSQLD, something --default-table-type that should be set to NDBCLUSTER.
AUTOMATIC FAILOVER ====================
8) From the study of the mysql cluster doc, I understand that NDBD nodes can be automatically restarted by the cluster itself, incase of failure of any of the nodes. My question here is, how can I configure the cluster for automatic failover to ensure that failed nodes are restarted automatically and recovered dynamically using node recovery protocol.
When a node restarts through StopOnError configuration variable it will automatically restart and perform the node recovery protocol.
9)When is the production version of MySQL Cluster going to be released.
4.1.9 was the first production version of MySQL Cluster, since then 4.1.10 and 4.1.10a has been released fixing some issues.
Rgrds Mikael
--------------------------------- Do you Yahoo!? Yahoo! Mail - Easier than ever with enhanced search. Learn more.
Mikael Ronström, Senior Software Architect MySQL AB, www.mysql.com
Jumpstart your cluster: http://www.mysql.com/consulting/packaged/cluster.html