Zero Downtime Nightly Backups: The Magic of Offline Checkpointing
If your backup and deployment strategies don't already take advantage of offline checkpointing, it's not too late to make another New Year's Resolution to git'r done this year!
Haven't heard of offline checkpointing? Well, it's not quite fair to say that you've been living in a cave, just because you're unaware of the wicked goodness of offline checkpointing. It's not like we put a major advertising campaign behind this little nuance of Perforce system administration. This is the sort of thing covered in KB articles and presented at Perforce user conferences. And I'm blogging about it now to help spread the word, because this particular nuance of Perforce administration has a big impact: No daily downtime.
Perforce Consultant Rusty Jackson is doing a Webinar on Offline Checkpoints on Wednesday, February 2, 2011 (Wednesday, February 11, 2011). See Details Here: http://info.perforce.com/offline-checkpoints
There is a good KB article on offline checkpointing that article gives the general gist. There are a variety of ways to implement a solution, including different variations of exactly what is meant by offline.
Offline can mean a second server machine. Bringing additional hardware into the mix allows you to extend the capabilities of your solution to better meet High Availability and Disaster Recovery requirements involving several servers distributed geographically. Offline can also mean just another directory on your main server machine. In that case, you trade additional disk space (roughly 2x the size of your db.* files) for zero-downtime backups. In some variations on the theme (including the one implemented as part of our "Server Deployment" consulting package), the server is taken offline for a few minutes per week to gain additional benefit of weekly reduction of database size. That "few minutes per week" is a maximum, regardless of the size of your databases. The exact time depends on how long it takes to replay just one day's worth of metadata into your offline databases. Large sites will generate more metadata, but tend to have correspondingly better hardware, hence the "few minutes."