We used to do the same in our 5.x environment. We decided given the new architecture with XI, we are going to give it a go with no health reboots to see if isolated service restarts are all it takes.
Coming from the legacy BO world, I initially proposed to have a weekly reboot of all servers. But as of this moment, this has not been implemented. We are in production for a few months and we have not experienced any problem whatsoever.
I have been recycling services every time when I rolled out a new release, and it has been quite frequent. In the first couple months, we were very aggressive and had weekly release.
Since we are still in phase 1 with limited number of users, this does not tell me much. When we move to phase 2 and have thousands of users, I will get a better feel.
We are running on 2003 SP1 with the CMS and data sources on SQL Server 2000 SP3. Our users only log in during normal business hours (in one Timezone) so I incorporate a recycle into the nightly Veritas backup. i.e. I have a Pre and Post script to shut down and start up the Database and all XI services every night at 23:00. This ensures we get a clean backup with no locked files - useful for a point in time restore. The disadvantage of this method is that if there is a tape/Veritas problem, then the backup job hangs and the services are still down in the morning. We have only implemented this nightly recycle in Development at the moment due to this very issue. However our Production services (which are not recycled every night) do seem to be stable. Be aware that there are Knowledgebase articles relating to Webi memory leaks - so make sure you research and apply the latest CHF if you are experiencing issues. Speaking if which…when is that CHF going to be updated on the BO site I wonder? ( http://ftp1.businessobjects.com/outgoing/CHF/boXIr2win_chf.zip ) I just downloaded and it is STILL the May 5th one. SP2 round the corner perhaps? Or maybe everyone is on holiday?
We are running Windows 2003 Enterprise Edition SP1 on all servers. Last December, we rolled out Business Objects to 3,500 stores with 20,000 users. Our users in Operation don’t stay on very long though. They typically just log on, get their reports and then log off. As of today, the highest number of concurrent users I have seen was around 400-500. It stays at around 70-100 through out the day.
So far I have not had a set schedule to recycle the servers. Sometimes, the servers are not recycled for a few weeks. I am quite impressed that the application is holding up very well.
Our next phases will be the rollout of workforce analytics, marketing analytics, and product analytics. Those users will stay on longer and perform more heavy-duty querying. I will wait and see how it goes.
There have been a few posts in BOB as to restart scripts - there is a bat one that utilises netstart and I have a perl script in BOB’s downloads that could be altered.
i.e. I’d suggest setting this up as scheduled job during your low peak hours.
Do a search on “sc.exe”, there’s a script that someone offered up which unlike netstart/netstop, allows you to stop/start the BO services across several physical machines. It’s a very helpful script in a clustered environment…
Sure - Service Monitor Script - Perl for Windows Services
Let me know if you need any guidance with it as it was more than just a restart script i.e. it would start up the services on a standby server if any of the primary node services would not start.