DB upgrade on Wednesday
Hello again, it's been a while since my last post here. I wanted to let you in on what we intend to do next wednesday during the extended downtime. The DB size has been ever increasing since launch and with increased size and number of players comes increased disk I/O. At the moment this is a performance bottleneck so we have purchased a FAStT 600 fiber storage server from IBM. This is a very scaleable solution and will allow us to grow well beyond the 120Gb limit the current solution imposes. Not to mention the performance increase which current estimates put at somewhere between six and tenfold. Unfortunately porting the db between disk solutions is easier said than done especially with a live system and hence the extended downtime. The way we plan to execute the migration is as follows:
On Tuesday we will be installing the fiber module cards into the sql servers. Since this is a cluster we can do this without any downtime required, we will simply install in one machine, fail the cluster over during downtime and then install into the other machine.
On Wednesday we will do an offline backup of the db, bring the new fiber box online, partition and format, rearrange the drive letters and move the db data and logfiles over. We are also considering splitting the db up into several filegroups on separate drives to increase performance. Time is the limiting factor for that operation so depending on the success we have on Tuesday we may or may not do that.
If all goes well the cluster will be up and running again on a new much faster storage solution at two o'clock on Wednesday.
We are also using the downtime to bring 12 additional sol servers online, rearrange both electrical and network cabling, bringing extra power feeds into the cabinets as overloaded circuits were determined to be the cause for the two power failures we have had since launch.
So we are flying to London early tomorrow morning and I will let you know how the preparations go.