>... It
>seems to have been designed with the right client/server attitude that
>we are looking for in order to backup our huge site (300 clients,
>probably around 200 GB/day).
FYI, I back up about 1/3 that many clients, but 3-4 times that much data.
I.e. Amanda works fine in large environments.
>For instance, Amanda's tape-changer interface seems to be geared towards
>the type of changer that has no idea what tape is in what slot. Amanda
>seems to just want to grab a tape, put it in the drive, and read the
>label, rather than simply ask the library which slot holds the
>particular tape Amanda wants. Is this the case?
That may be the case with the changers that come with Amanda. I wrote
my own (it's pretty easy) and my "slots" are VSN's that I ask the library
software for directly.
>If I understand things correctly, the only way I can get Amanda to use
>multiple drives in parallel would be to create multiple "instances" of
>Amanda and give each one a different tape drive. ...
Yes. Or the changer can ask for "any tape drive" and the library software
picks an appropriate one (that's what mine does) so you don't have to
dedicate a specific drive.
Note that multiple "instances" (Amanda calls them configurations) may
run simultaneously on the same tape server machine. I used to run four
at a time on a machine with a large robot until the network bandwidth
needed got out of hand.
>This seems like a
>difficult and error-prone method to use for a large site, trying to
>remember which clients are backed up with which instance of Amanda.
It works for me, but in my case, the client list and load per server are
pretty static, i.e. I don't have several new/deleted clients per week
or wildly changing dump image sizes. It's pretty much a set it up once
and then just keep an eye on it kind of operation.
There are plans to enhance tape (actually, image output) handling,
but they are way down the road.
>David DeSimone
John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
