Rt-3 cluster mysql backend

I am looking to migrate the backend (mysql) to a clustered platform and
was planning on using mysql 5.1.
However rt-3.6 need Innodb and the cluster needs ndbcluster … have
anyone had a work around for this … or possibly more general have
anyone implemented a mysql clustered backend …
On google I came across
:http://forums.mysql.com/read.php?25,32782,32812#msg-32812 …but unless
I am missing something the innodb tables would n’t be replicated in the
suggested solution …Any help will be greatly appreciated…
the system will be rt-3.6 / dbix 1.43 /mysql 5.1 /gentoo / apache2
/mod_perl2
Roy

Roy El-Hames wrote:

I am looking to migrate the backend (mysql) to a clustered platform
and was planning on using mysql 5.1.
However rt-3.6 need Innodb and the cluster needs ndbcluster … have
anyone had a work around for this … or possibly more general have
anyone implemented a mysql clustered backend …
On google I came across
:http://forums.mysql.com/read.php?25,32782,32812#msg-32812 …but
unless I am missing something the innodb tables would n’t be
replicated in the suggested solution …Any help will be greatly
appreciated…
the system will be rt-3.6 / dbix 1.43 /mysql 5.1 /gentoo / apache2
/mod_perl2
Roy

Last time I looked, NDBcluster still required to be able to fit the
complete DB into RAM (apart from the other problems).
How large is your database?
Also, ISTR Jesse saying that the DB is not really the bottleneck any
more (in most cases) - years after years of optimizations have left
their traces…

What are your requirements regarding the availability?

cheers,
Rainer

Rainer Duffner wrote:

Roy El-Hames wrote:

I am looking to migrate the backend (mysql) to a clustered platform
and was planning on using mysql 5.1.
However rt-3.6 need Innodb and the cluster needs ndbcluster … have
anyone had a work around for this … or possibly more general have
anyone implemented a mysql clustered backend …
On google I came across
:http://forums.mysql.com/read.php?25,32782,32812#msg-32812 …but
unless I am missing something the innodb tables would n’t be
replicated in the suggested solution …Any help will be greatly
appreciated…
the system will be rt-3.6 / dbix 1.43 /mysql 5.1 /gentoo / apache2
/mod_perl2
Roy

Last time I looked, NDBcluster still required to be able to fit the
complete DB into RAM (apart from the other problems).
How large is your database?

The db is around 18G and Growing … (average 1200 tickets a day) … as
far as I know NDBcluster require just the indexes in memory …I might be
wrong.

Also, ISTR Jesse saying that the DB is not really the bottleneck any
more (in most cases) - years after years of optimizations have left
their traces…

What are your requirements regarding the availability?

Its not really a question of availability as much as performance … in
my case the bottleneck is the db … (the db server is dual cpu 2.8G and
6G memory) …
the web server is very much similar spec and the performance graphs
showing no signs of stress , where the db server is in most cases at
peak (cpu and memory) …
I have applied every patch Jesse have recomended and to be fair there
isn’t a single query that I can blame …
What do you recommend ?
Roy

Roy El-Hames wrote:

Rainer Duffner wrote:

Roy El-Hames wrote:

I am looking to migrate the backend (mysql) to a clustered platform
and was planning on using mysql 5.1.
However rt-3.6 need Innodb and the cluster needs ndbcluster … have
anyone had a work around for this … or possibly more general have
anyone implemented a mysql clustered backend …
On google I came across
:http://forums.mysql.com/read.php?25,32782,32812#msg-32812 …but
unless I am missing something the innodb tables would n’t be
replicated in the suggested solution …Any help will be greatly
appreciated…
the system will be rt-3.6 / dbix 1.43 /mysql 5.1 /gentoo / apache2
/mod_perl2
Roy

Last time I looked, NDBcluster still required to be able to fit the
complete DB into RAM (apart from the other problems).
How large is your database?

The db is around 18G and Growing … (average 1200 tickets a day) … as
far as I know NDBcluster require just the indexes in memory …I might
be wrong.

According to:
http://www.onlamp.com/pub/a/onlamp/2006/04/20/advanced-mysql-replication.html

that is a feature for 5.1

And besides, NDB doesn’t seem to support foreign keys, which pretty much
rules out its use for RT.

Also, ISTR Jesse saying that the DB is not really the bottleneck any
more (in most cases) - years after years of optimizations have left
their traces…

What are your requirements regarding the availability?

Its not really a question of availability as much as performance … in
my case the bottleneck is the db … (the db server is dual cpu 2.8G
and 6G memory) …
the web server is very much similar spec and the performance graphs
showing no signs of stress , where the db server is in most cases at
peak (cpu and memory) …
I have applied every patch Jesse have recomended and to be fair there
isn’t a single query that I can blame …
What do you recommend ?

My DB is still much smaller (and I use PostgreSQL anyway).
IIRC, other people with lots of new tickets usually delete most after
some time.

cheers,
Rainer

Its not really a question of availability as much as performance … in
my case the bottleneck is the db … (the db server is dual cpu 2.8G and
6G memory) …

And you’ve spent time optimizing the my.cnf?

FWIW, the problem with ndb_cluster, last time I looked, was transaction
support. RT doesn’t need innodb. It needs the ability to isolate and
roll back transactions.

Jesse

Is there any chance that RT will ever have active/archived tables
or databases instead of growing the active ones boundlessly?

There’s a chance of anything if we see good patches and/or big fat sacks
of cash.

Jesse