On Thursday 06 February 2003 01:01, John Andersen wrote:
On Wednesday 05 February 2003 09:55 pm, Travis Owens wrote:
Try Arkeia -- http://www.arkeia.com It's won A++++++ awards every time it's reviewed,
Yeah, A++++ reviews till you need to restore from a crash. Then you are Out of Luck unless you have some OTHER backup to restore the tape catelog from. [snip]
This is not true. The tape cataloge can be restored from the tapes themselves. I have used Arkeia for at least 5 years. Had great experiences using it. It is a great product. I used it at Caldera where I was a system admin/Software Engineer (infact, I worked with several of the German engineers that now work for SuSE). I tranined Caldera's IT deparment on Arkeia's use and SCO now uses it for company-wide backup. VTI, an SCO spin-off company uses it to backup their company infrastructure. I used it at the University of Utah while I was a system administrator there. I convinced the University of Utah's Center for High Performance Computing to use it and they have now for the last 5 years for their Beowolf cluster (configuration and such) backup as well as workstations and other UNIX server backup. They backup several servers and workstations of varying OS types. It has worked great for them over the years. Like any industrial strength product, there is a _lot_ to know about Arkeia but you can start using it without knowing everything about it. It is very powerful and flexable. It supports encrypted transfer of data between clients and the server, simultainious streaming from several clients. It is strait out, the fastest client-server backup product I have used on Linux. The header of each tape containes the C-source for a command line restore tool. This is also included in the software. There is a command line tool for extracting/restoring the tape index from a set of tapes. I have personaly used that tool and know that it works. A downside is that in versions < 5, the tape index database is filesystem based. Three files are created for each file that gets backed up. In my case at Caldera, I was backing up almost 1.5 Terrabytes of data in a full backup. Three files for every one that I backed up makes a machine very slow just for database access. One should tripple the ammounts of inodes on the partition used for the database. If lost, the database also takes time to restore so my advice is to copy/tar the index to another machine after each backup. There is better handling of the database in the latest version from what I understand. There are some tricks to make dealing with the database better. For example, people have created a file, mkfs.whatever (usualy riser or ext3) on the file and moutned it as a loopback deviece for the database. Mounted this way, the database seems to be extreamly fast and can be backed up by just copying the one file, and not dealing with accessing individual files within the database. This should only be considered if you are in a large environemnt, backing up a lot of data. Most people will never need such a work around. That being said, it's not for everyone. If you're only backing up one machine, use something else. If you like having the source code to the product, use something else. Use the right tool for the right job. -- Marc Christensen http://www.mecworks.com/ http://www.mecworks.com/~marc/resume/