The success of the hosted offer of InDefero is above my expectations. This means that my strategy to backup the data is not any more the right one. At the moment, everyday a full backup of each forge database (in fact PostgreSQL schema) is performed, each subversion repository is dumped and everything is backed up in two remote data centers.
Basically, once a day, a full backup of everything is performed and copied in two different places. This also means that you can lose up to 24h of work in the worst case scenario (a crash just before the backup). I must say, I am not comfortable with this 24h window.
So, I am pleased to announce the start of the work on a new strategy (which will run in parallel of the current one for a while to test it). The strategy is to have a warm standby server ready to jump in to take over the hosting in case of failure of the main hosting. The good points:
- the backup server will be in a different data center controlled by a different provider;
- the data will be split in two sets, database and repository data, stored in two different RAID6 filers;
- the sync of the data will be performed every 5 minutes for the database and I hope in near real time for the repository data.
Basically, once setup your data will be stored that way:
- main hosting on a server RAID1 storage for both repository and database;
- backup in a different datacenter on two RAID6 filers (one for the repositories and one for the database);
- near real time backup from the main server to the backup server.
Keeping your data is a continuous process and sadly it is a process where at any time something can go wrong.