On Tue, Sep 29, 2020 at 10:55:51AM +0700, Olivier wrote: > nice thing part of Amanda is written in Perl and integrates well with > some admin scripts here and there).
Quite nice indeed! I wasn't aware of that detail, but our IT group is very Perl-centric, so good to know we're well-situated to make changes or extensions to amanda if needed. > One of my concerns would be the network bandwidth, you'd need at least > 10Gbps network to push that amount of data through in one day. That is > where Amanda trying to balance the size of the backup across a cycle may > come handy. I think we should be able to manage it. Available bandwidth is sufficient for the current TSM solution through the law school. The two largest servers typically take 6-8 hours to complete their backup runs, so a bottleneck at the amanda server itself could be an issue, but I can probably get campus data services to give us a faster link for it if needed. > I think that independant disks is better than any RAID thing that > would either waste some storage, or render the array unusable if any > one disk gets faulty. If we go disk-based/vtape for storage, I'd be looking at RAID 1 for protection against disk failure (which we've had trouble with in production systems in the past). Definitely not going to use anything that would increase exposure to data loss, of course. > I have 8 disk bays, one disk for system, holding disk and a copy of > the final state of the accounts about to be deleted. Another disk for > some additional holding space and a second copy of the deleted > accounts and the 6 other disks for vtapes. Are those hot-swappable? I've made a couple half-hearted attempts at hot-swap disk in the past, but never managed to make it work myself. > Amanda is not very well suited to backup workstations when mot of the > users will turn off their machine after work. I don't know if that is > part of your plan, Nope, I'm not backing up any end-user workstations. Glusterfs storage cluster, some standalone Windows-based and FreeNAS storage servers, half a dozen kvm host servers, and a bunch of (mostly-)virtual servers providing actual end-user services. > I have been looking at some solution that would run on the user's > workstation and that would push a backup on some server and use Amanda > to keep a backup of that server. I've got a couple developers who are backing up their workstations via rsync to the FreeNAS server, which sounds like basically the same concept. They seem happy with it, although we've never needed to do a restore on any of that data. > One think I implemented a long time ago is a replication of Amanda > configuration and indexes: at the end of each dump, I push a copy of > that data with rsync and also send that information to me by email > (email being replicated automatically on the email server). With the > disks and the indexes, I can manually extract about any backup from a > machine that would not even have Amanda running. I choose to use tar > rather than dump for that compatibility advantage. Seems like a good tip. Thanks! -- Dave Sherohman
