J. Shirley wrote:
On Wed, Apr 9, 2008 at 8:22 AM, Clinton Goudie-Nice
<[EMAIL PROTECTED]> wrote:
Our database looks the same; however since you can only have one master
tracker running at any time, no one will be running updates against your
other master DB.
Clint
Just out of curiosity, have any production installations attempted to
run on a MySQL Cluster (NDB Cluster)? It seems like a slightly safer
method for scaling upwards without having to deal with a 3rd party HA
VIP setup.
-Jay
ndbcluster not production ready yet.
We tried to switch from master-slave to ndbcluster.
Boosted our performance quite some.
But at high load nodes eat up memory and start tangling.
so basically for us we put ndb online ... let the portal connect and
first 30seconds i performed awesome.
Then nodes used up memory by replicate to each other and started to fail
... resulting taking down the whole cluster.
(leaving it in an unstable, not recover able state .... means another
5hours to restore a backup)
(setup was 8 db-nodes with each 4*4GB RAM 4*8GB RAM)
using ndb on test setup though ... runs stable but it uses quite some
memory.
Basically:
when you have many changing files you need enormous amount of RAM on
the db nodes.
when you have many files in whole you need enormous amount of RAM on
the db nodes on reserve or ndb locks that table (config file stuff and such)
when you only have about 5millon files and they don't change ... no
need for ndb ... just stick to master-slave by adding slaves to mogadm
or loadbalance the over lvs.
lg
