Forum OpenACS Q&A: Response to little question about a server

Collapse
Posted by Malachi Kenney on

Finagle's Law applies in full force here. In my experience, switching to RAID later usually means switching to a different, RAID capable system rather than upgrading a live system to RAID. The former method (a standby system) is both faster and safer than trying to upgrade a single production box.

The procedure for upgrading the live system will mean full backups to different media and a complete restore after the upgrade. You can, thanks to Finagle's Law, count on the downtime between those two operations being measured in hours rather than minutes unless you have practiced the drill with that specific hardware in advance.

An important factor here is that the kernel configuration bears very little relation to this operation. You can indeed configure the kernel beforehand. You can not, however, entirely configure the hardware RAID and the filesystems (especially if you plan to use the single drive in the final RAID configuration). As an example, adding drives in to the machine and then initializing a 3 drive harware RAID 5 with the standard Dell Perc tools may take 40 minutes of complete downtime before any actual migration takes place.

If you know you need hardware RAID then it is simpler to go with that configuration from the get go. Otherwise, unless you really know what you're doing, you should plan on non-trivial downtime.