Multiple Issues - Project Server 2003 SP2a

[Colin]

Our setup -

6 boxes
2 x Web Server in NLB cluster (one host currently STOPPED)
2 x App Servers (SM and VP) in NLB cluster (one host STOPPED)
1 x 2 node SQL Cluster (Active Passive) with AS on the 1st node

All servers are Windows 2003 Server Enterprise Edition SP1 + security
updates etc.

Issue 1
XML files on the VP server are not being deleted, although it appears that
some/all of the changes are actually getting through to the database and
being shown in PWA instantly. Update: just been told they have started going
through again, very weird.

Issue 2
Occasionally we get PWA hanging when attempting to log on (most users at
once). When this is happening there are no obvious CPU, MEM or Disk issues on
any of the servers.

Issue 3
We occasionally get CPU "race" conditions with SQLSERVER.EXE which flatlines
at 50% CPU. If I stop start SQLServer it goes away. Memory and disk all
appear ok on the SQL box.

We've had PSS calls open for as long as I can remember (since
implementation) and we're not sure this has even been implemented correctly,
we're lacking confidence in the 3rd party that did the work shall we say.

I've re-run the complus tool just in case that would help, doesnt seem to
have.

We have the relevant SQL re-index jobs setup, so I'm told by the DBAs.

Our domain account used for EPM is not locked and is defined with the
correct password in all services.

Another thing that may be of use, we used NETWORK AUTHORITY when setting up
the connections to the database from the front ends, the 3rd party that
installed the software hadnt done it that way before. We have done it that
way in our previous installation and that system was in fine working order.
The theory here is that the clustering has caused some config or bug issue in
the background. Hence why we've stopped one each of the web and app servers
to reduce that possibility. (well the AP server has to be stopped as the SM
doesnt like being clustered).

PSS have told us to re-install the whole system, which we may have to. I
just wanted to see if I could get some pointers as to where to look for each
of the problems above.

 

[Scott m Wagner - EPM Consulting Services]

Colin,

There should be no issue with load balance on the front end web servers. I
would recommend putting Views processing and Session Management services on
separate boxes. These services are mostly single threaded so load balancing
helps very little and often poses many issues.
I also am not a fan of clustering the database servers as I have had many
issues. I would possibly recommend separating the databases over the two
servers rather then clustering them.
To elevate down time caused by failure I would ensure backups of the systems
to an external (Outside the solution) source.
2X Web Server Front NLB - PS 2003 - Windows Server 2003, SharePoint If you
are using it.
1X Session Manager Server
1X Views Processor
1X Web Database Server (Project Databases) - *
1X (Project Database) SharePoint Databases - *
*- The databases will depend on if you use SharePoint and how much it is
used. In larger organizations I see more use on SharePoint Database Server
than on the Project Database Server so it really depends on usage and
configuration.

IMHO - This would be the best configuration for a 6 server solution.
I believe the issues with the XML files not processing are due to the
configuration. Load Balance on that server.
I hope this helps, it you have additional questions please feel free to
email me direct. (e-mail address removed)

 

[Colin]

Scott,

thanks for the reply, I have emailed you with the current situation. I'll
post here for reference too.

The management have decided to move the system to the US (from the UK) and
they are pressuring us to reduce the number of servers as much as we can.

In light of this I have suggested they source 3 servers and put the web
interface on one, VP and SM on another and the SQL DBs on the other.

If we have 1400 global users (accessing via the Citrix farm), concurrency is
low - maybe 20 project managers max, with possibly 200-300 time sheet users
on a busy day, does that sound acceptable?

The servers will likely be 2 x CPU (3Ghz +) with 4GB ram etc. HP Proliant
DL380s I'm guessing at the moment though as the servers have not been
sourced.

Any comments?

Will we continue to see problems with this setup?

I believe that we wont, as I think most of our issues are #1 to do with the
NLB setup and #2 the 3rd party that installed didnt really know how to
install the system, so their may be other errors.

Any comments will be appreciated.

Regards.
Colin.