We have been planning a migration to better and faster (and more) servers for a while. Adding new servers is quite easy, but when you finally need to cut off old expensive servers it gets more complicated. You must ensure no data is lost, update all DNS settings and route traffic from old IP addresses to new ones.

From our last crash we have learned that data security is not the only very important thing we should care about at Wikidot, but also performance and availability. We added many monitoring processes and statistics to the service. This allowed us to see what performance bottlenecks we have and how to deal with them. We also started to create a highly-redundant hardware infrastructure, that allows us to deal with the most common (and some of the least common) hardware problems that can occur.

From our experience, the most faulty piece of computer hardware is hard disk. We have already replaced many of them. This is why we always use RAID to be safe from disk failure. But as other components are much more reliable, you can never fully rely on them.

This is why we are creating a cluster of servers with data mirrored across them and each of them being approximately the same machine that can handle any of the functions our servers have (e.g. application server, database, file and DNS servers, and the e-mail sending machine). Moreover each of the servers in this cluster is powerful and reliable enough to get the additional jobs done when one of the servers fail, while still coping with its original functions quickly and efficiently.

Yesterday we finalized the migration of the wdfiles.com service (serving files uploaded to Wikidot sites, code block fragments and recently introduced [[html]] fragments). Site backups are now properly done and renaming of files works again.

If you still have any problem with files, extracting code (or html) blocks or logging in to private sites please let us know.