Grouper Users - Open Discussion List

[Andrew Morgan <>]

On Fri, 20 Feb 2015, David Langenberg wrote:

> On Thu, Feb 19, 2015 at 4:36 AM, Mandeep Saini 
> <>
> wrote:
>
> >  Dear fellow Grouper users,
> >
> >
> >
> > We have started to get acquainted with the Grouper very recently and
> > currently in the process for setting up our test, UAT and production
> > environments. For the same I have few questions:
> >
> >
> >
> > -        What is the recommended capacity dimensions for Grouper VM in
> > terms of CPU/RAM/Storage?
> We're running dual cpu 2.67 GHz VMs with 6GB ram.  This is an older install
> though, so I expect the RAM to go up in our next iteration.  Storage on the
> grouper servers isn't really a concern as you just have to hold the logs.
> The Database is consuming about 50 GB with the majority of that space going
> to the audit logs and changelog.

We are bringing up a new Grouper v2.2.1 installation here at OSU.  Our 
backend server has dual 2.3GHz cpus with 4GB of RAM.  Our Java heap size 
is set to 1.5GB right now.  You may need a bit more depending what your 
loader jobs look like and how many provisioners you are running.  Either 
way, the requirements are not extreme.

We are running a pair of frontend servers for the UI and WS components. 
These are behind a hardware load balancer for redundancy.  Each frontend 
VM has dual 2.3GHz cpus with 4GB of RAM.  The Java heap size for Tomcat is 
really pretty small right now (256MB).  We'll see if we need to increase 
that as we ramp up usage.

> >  -        Does anybody have experience of using SQL server with Grouper?
> > If so what was your experience? I am asking this because in one of the
> > training videos I saw Grouper team recommends MySql, PostgreSQl or Oracle
> > and not SQL server but we want to use SQL Server purely for HA reasons as
> > we have SQL cluster already setup and would like to reuse it if possible.
> Everybody who's tried to use SQLServer for their grouper DB pretty
> universally regrets their choice.  Performance is horrible.  If you really
> do want to use it, though, the project would appreciate any tips, tricks,
> and pull-requests for fixes of SQLServer performance problems.  That said,
> Chicago does have the grouper-loader attached to a SQLServer database and
> finds the performance from that standpoint to be acceptable. We have also
> played with the idea of using that SQLServer instance as a subject source,
> but still have a lot of optimization to do before that's ready (it's slow).

I'll second what David said.  We did performance comparisons between 
SQLServer, Oracle, and MySQL.  SQLServer was terrible, and we couldn't 
really figure out why.  Oracle and MySQL performance are pretty similar 
(we opened some bugs on MySQL performance that were fixed in v2.2.1).  We 
decided to go with Oracle because we have an existing Oracle RAC cluster 
that the Identity department doesn't have to manage!

        Andy