[389-users] VM Experiences

[Andreas Andersson]

Hi!

I have some questions regarding experience with using production 389  
DS / RedHat DS in a production environment.

Some info about our environment:
I'm running an environment consisting of two masters in a multimaster  
setup replicating to eight consumers.
Our database is divided into several backend databases which results  
in a total of 40+ replication agreements.

Our load is approximately around 24000 returned entries a minute and  
we try to keep all entries in the entry cache.
We are thinking about virtualizing parts of this environment and I'm  
wondering if anyone on this mailinglist has experience with  
virtualizing whole or parts of a production environment.
Database size: Around 350 000.
Entries returned: 400 entries / s per replica.
Connections: 30 / s per replica (total 1000+ replica)
Current hardware: Blade servers, RHEL 5, 8gb memory.

Previous experience:
We have tested to run larger environments in virtual machines (VMWare)  
in our lab which resulted in replication errors. This was a couple of  
years ago and related to timestamps. Consumers with times older than  
the replication, master, timestamp. Masters halted and other  
replication errors worried us.
We were of course using ntp and there was some kind of fix for the  
time shifting.
Another thing that comes in mind was "hickups" that was pretty common  
in our load tests.

It would be very nice to hear other experiences with using 389 DS in  
larger production environments.
Everything from replication stability to performance figures comparing  
using hardware and virtualized machines.

Regards - Andreas