We need to set up a database system (flat files only) for international members with the need to make these databases inaccessible per time zone for about half an hour between 2 am and 2.30 am every night for every time zone, for maintenance purposes.

It seems we need 6 disks holding a number of time zones each and possibly 2 time zones needing 2 disks each or one very large disk (>300G). This has to be doubled for raid 1 and trebled for a hot swap.

The question we have is how to set this up for a hardware raid 1 environment and be able to automatically disconnect every time zone (24) at the correct time. We're using Debian Sarge 3.1 and Apache2 although it seems, may be, we should use apache 1.3 instead.

Thank you for your views.

With software RAID, you could pull a partition or DRIVE out ot the array at the appointed time, treat it as a snapshot, do maintenance, put it back in the array and let it re-sync. This would require no interruption of service. If you have a commandline/script interface to the hardware controller, you may be able to do this with hardware RAID, but I doubt that would work with the GUI interfaces I have seen. If mainenance is more than a backup, it gets complicated because you will want to merge data from the wandering drive, with the stationary drive. Staying coherent would require intelligent software.<p>
If you have a way of breaking connections safely, you could just change file permissions on the affected directories to allow access from the maintaining user only.

Thank you for your answer.
The maintenance involves us update the member files as well as backup and we need the system automated.
We would prefer very much to have the system unavailable for a short time rather than exposing ourselves to the lack of raid 1 while one drive is removed for backup.
I don't know much about raid and I was tempted to think that hardware raid was more versatile than software raid, your answer seems to indicate the opposite. There are those hardware cards that are really software raid to complicate matters.
We were planning to use a true hardware card like 3ware 9500. The system would have to deal with a fair bit of encryption and I wonder if adding software raid would be too much although I have nothing against implementing a software raid if that's the solution.
Is it an option to use rsync to keep a hot swap (or similar) as a ready backup (ready to be removed and saved)?

"Is it an option to use rsync to keep a hot swap (or similar) as a ready backup (ready to be removed and saved)?"
-------
That should preserve the data but I do not think you could just put the rsynced volume back in the RAID array. Magic numbers and all that... What about doing RAID over the LAN? If your system is not too busy, and the link is gigabit/s it could keep up. See

http://www.it.uc3m.es/~ptb/nbd/#Introduction This would have the advantage that your backup function would already be done when you interrupt service. The update is likely quick and you could be back on line sooner.

The original question of how to handle the interruption remains. You could reload the server configuration to deny remote access or you could change the routing or file system to give a polite message. This could be rude treatment for existing connections. Some sort of continuous mode of operation would be better. If your system is truly multi-user, perhaps the file locks it has would permit the updates to occur during normal operation.

It looks like disconnection of the system for maintenance is not a good idea and more difficult to implement than I envisaged, so I think we'll follow your idea to keep the system available continuously. The maintenance still has to be implemented so that's no problem.

I was unaware of NBD and RAID1 configuration, I've read the article you mentioned and another I found but I can't figure out what benefit there is in it compared to plain raid1 with 2 disks on the same machine. Is it the fact that if one machine misbehave, the other has the necessary data?

Thank you very much for your advises, this will probably be my last question.

Yes, RAID over LAN gives you two of everything so you get greater redundancy. The two parts can be in separate locations to preserve data/operation in the event of serious tragedy. Gigabit/s is good to 100m so separate buildings are feasible. Power supply/drive/mobo failures will not cause loss of data. You can route things so that both boxes can serve in a fail-over mode.