Software Upgrade Considerations
We can't stress the importance of keeping your application and infrastructure software up to date. Many companies are moving to a network-deployed push model for software upgrades and fixes. This guarantees that all connected components in their infrastructure (servers, routers, workstations, and so on) are as up to date as possible at all times. If you are in a 24x7x365 uptime requirement for an application, finding down time to upgrade or apply fixes might be a little difficult. It is less difficult if you are using a multi-server clustered configuration that permits you to take a server offline (out of the cluster), apply service packs to it, and then rejoin the cluster. Keep these types of options in mind when making that final high availability solution choice. Here are the basics:
Plan on downtime (if possible)— This can usually be negotiated during your end-user SLA definition time. The value of planned downtime for software/hardware maintenance is huge. You don't need much time, but plan for it. The planned downtime doesn't count against your high availability percentage performance measurement.
Be aggressive about applying upgrades— If possible, use a network-deployed push model to periodically upgrade servers and workstations. Or, selectively plan on at least monthly upgrades as part of your planned downtime maintenance (if you have planned downtime). This also covers application software, middleware, and database software. Some of these won't have frequent updates or upgrades, but you must stay current on these. If you have problems with any of these software components, you will get better support from the software vendor if you are on the latest release.
Antivirus software— Make sure you have antivirus software in place and it is active at all times but probably not have it scanning your database files due to the performance impact. (I'm not sure there are any exceptions to this rule any more. Even at the time of this chapter's writing, a major “attack” virus has been hitting servers and workstations around the world.)
Close backdoors— Make sure you have closed all backdoors (ports, and so forth) that are not being used, make sure you have some type of firewall is in place (hardware or software), and turn off things such as anonymous FTP accounts.
Configuration snapshots— Make periodic snapshots of the configuration you have and the release levels of all software components in our high availability solution (you can use SQLDIAG.EXE for this if you want). Suggest this be done at least twice a year, if not quarterly. Be proactive here, it will also keep you honest on all software upgrades that are needed.
Automatic Updates— Microsoft has an option to automatically update Windows but we recommend that this be turned “off.” So, leave on the notification of new updates, but turn off the automatic installation option. You need control over what update to apply and when to apply it.
Vendor agreements— Get all vendor agreements in place in the form of current software licenses, renewable software support agreements. This should include a service level agreement with the vendors that corresponds to your service level agreement that you will have with your end-users (and matches your high availability requirements).
All of these items have a direct effect on high availability. Consider them carefully.