jrg@apple.com (John R. Galloway) (04/26/89)
One of my current clients has a large network of Suns, diskless and diskfull 3/50's, 3/60's, 386i's, NFS servers and of couse ND servers. Their current policy is to shutdown the NFS servers for nightly backup. I find this somewhat annoying since then my diskless node is basically (if not completely when the ND server, which is also an NFS server is down) useless. The rational seems to be that if not shutdown the backups are indeterminate, if you want to restore you can't tell for sure what tape to uses since a file (particularly one in an ND partition?) might or might not have been backed up. It seems like some sort of network administrative tool is needed here to manage backups. The situation is even worse if you include managing the backup of diskFULL workstation in a uniform way withOUT the need to bother the stations user (i.e. by the MIS dept.). What exists along these lines? How do other large sites handle these issues? apple!jrg John R. Galloway, Jr. contract programmer, San Jose, Ca These are my views, NOT Apple's, I am a GUEST here, not an employee!!
david@indetech.uucp (David Kuder) (05/11/89)
In article <28358@apple.Apple.COM> jrg@apple.com (John R. Galloway) writes: One of my current clients has a large network of Suns, diskless and diskfull 3/50's, 3/60's, 386i's, NFS servers and of couse ND servers. Their current policy is to shutdown the NFS servers for nightly backup. The rational seems to be that if not shutdown the backups are indeterminate, if you want to restore you can't tell for sure what tape to uses since a file (particularly one in an ND partition?) might or might not have been backed up. There is a long running discussion about wether "dump(8)" can be run on active disks reliably. At the USENIX Large Systems Administration conference large amounts of anecdotic evidence that running dump against lightly used disks was fine. The concensus was also that you pay your money and take your chances. The general compromise was to run daily incremental dumps at times when little activity was expected and full dumps against shutdown disks. In article <28358@apple.Apple.COM> jrg@apple.com (John R. Galloway) continues: The situation is even worse if you include managing the backup of diskFULL workstation in a uniform way withOUT the need to bother the stations user (i.e. by the MIS dept.). What exists along these lines? How do other large sites handle these issues? The Large Systems Admin. conference had several papers describing different schemes. The most common was to do disk to disk backups from the workstations to the server and then do backups from there to removable media (tapes or disk packs). One site used a "networked tape ape" who ran around from WS to WS sliding in cartridge tapes for the nightly cron dump. ____*_ David A. Kuder {sun,sharkey,pacbell}!indetech!david \ / / Independence Technologies \/ / 42705 Lawrence Place FAX: 415 438-2034 \/ Fremont, CA 94538 Voice: 415 438-2003