[376] in Info-AFS_Redistribution
No subject found in mail header
daemon@ATHENA.MIT.EDU (NIK%ZURLVM1.BITNET@pucc.princeton.)
Thu Nov 7 10:40:52 1991
Date: Thu, 07 Nov 91 15:32:57 SET
To: Info-AFS@transarc.com
From: NIK%ZURLVM1.BITNET@pucc.princeton.edu
Date: 07 November 1991, 15:32:26 SET
From: Michael Niksch 0041-1-7248-913 NIK at ZURLVM1
To: Info-AFS@transarc.com
Subject: AFS backup system
I am desperately trying to set up the AFS backup system, and the
more I try, the more problems arise. I am running rs_aix31.
What I actually want to achieve is an automated backup scheme that
clones all volumes every night, then does a full dump to tape once
a week, and incremental dumps every day. The important constriction
is that we want to manually change tapes once a week at most, and
then all the tapes at a time. We DO NOT WANT to put in a tape, wait
for a dump to complete, then put in another tape, and so on. We are
currently condsidering up to 20 GB of data, and we are more likely
to get additional tape drives than to get any human operators.
The backup system should mail its logs to the administrators,
asking for intervention only if something fails, like a wrong
tape being mounted or a volumeset exceeding space on the tape.
First of all, backup and butc want to run interactively.
Somebody suggested to me using EXPECT, a package designed
to automate interactive applications. Well, I managed to
do something similar, running backup and butc as coprocesses
and watching the pipes they are connected to from a shell
script. I got it to work for readlabel, scantape, and dump.
The script is simply based on heuristics, as there are
apparently no rules as to what message is printed by butc,
what message is printed by backup, and what goes to stdout
and what goes to stderr. Getting the output buffers flushed
to the pipes in the place of ttys is yet another problem.
Anyway, now I am trying to implement a backup scheme using
multiple tape drives on multiple machines. According to the
AFS manual, I was under the impression that one can have
up to 7 (what magic value is that) drives per machine,
and use any number of backup servers. However, looking
at backup addhost and /usr/afs/backup/tapehosts, I learn
that one backup server knows about the others' tape drives.
If I run backup status <n> on server A, it indeed seems to
connect to butc <n> on server B if tapehosts lists it this
way. The next problem is in backup add/del/listhost. I tried
to configure 3 tape drives on a machine from the backup prompt,
and listhosts would happily list all of them. Now I quitted
backup and called it again, and guess what - it knew only
about the first and last entry in tapehosts. The same seems
to hold for any number of tapes - only the first and the last
entry in tapehosts are read when backup starts up. If I do
an addhost, backup will rewrite the file and all entries but
two are erased from tapehosts.
To summarize my questions:
1) Is anybody out there ACTUALLY USING the AFS backup system at all ?
2) Has anybody managed to automate it enough to meet our requirements ?
3) If so, what amounts of data do these systems handle ?
4) How many tape drives can you use per machine / per cell ?
5) Has anybody tried to use a tape robot ?
6) Does the AFS backup system compress data when dumping volumes ?
7) Would you recommend to just use vos dump, then handle the resulting
disk files by some other backup and tape control system ?
8) If so, can you do incremental backups, as incremental vos dump is
reported to be broken, and incremental vos restore does not exist ?
9) Would you recommend to just run tar on the AFS filesystem with a
system:administrators token ?
10) Is anybody out there running AFS as a service (not just a playground)
for 200+ users without regularly fixing and recompiling AFS source
code himself ?
11) If not, how far would you get with the export version (no DES license)
and without an AIX source license ?
Michael Niksch
