Not really - the only thing I can think that we ever reboot systems for is if we are doing a kernel upgrade, a physical hardware change, or something of a similar magnitude.
Web server upgrades should be near hitless - even if you were running a single server environment. Luckily, I have 10 clusters globally, so performing upgrades doesn't really cause an issue.
Database access is the killer for me, as each cluster reports into its local database managers, and then these get pooled on request for reporting data etc...
So ensuring connectivity between all the clusters is vital to the reporting functionality.
You see a large change with the big corporate that not only require the service to be functioning 100% of the time, but all ancillary services such as reporting, configuration changes etc... to also be functional near 100% of the time. We've actually just bought in an SLA covering reporting availability, and have had an SLA covering configuration changes for quite some time. The primary service has always been covered by an SLA - however, as with all SLA's, one of the most important things to have in place before you commit to them is a way that you can measure and ironically, in the case of the reporting SLA, report on whether you are succeeding, or failing and where you need to improve.
|