[216] in Athena_Backup_System
improved support for crossing tape boundaries
daemon@ATHENA.MIT.EDU (Diane Delgado)
Thu Mar 21 14:46:12 1996
To: athena-backup@MIT.EDU
Date: Thu, 21 Mar 1996 14:46:01 EST
From: Diane Delgado <delgado@MIT.EDU>
Hi,
Athough it won't happen often in our environment, it seems that
since we are allowing dumps of a single dumpset to span tapes,
we might want to (eventually) provide better support for this. For example:
- log tape info after each tape is complete, instead of after
the whole job completes. This becomes very important if
we want to support stackers since the slave would want to be able
search the stacker device for a useable tape. We would
want the tape which the slave had just written (in the case of a dump
which spans multiple tapes) to have been logged with
the master so that it would show up as unavailable for writing.
It is possible, though, that during the portion of the dump which
writes to the second tape that there were too many severe
errors (e.g., fileserver crashes during middle of dump).
Does anyone see any issues with allowing some tapes of a multi-tape
which are good to be kept in the database as opposed to the
all or none scheme which is currently implemented?
- Would we want to eventually support the ability to resume a dump if one
of the subsequent tapes, say the 2nd tape, has too many media
errors and needs to be replaced? (This functionality would allow
the operator to replace the bad tape and the Slave would restart
the dump at the point where it started using the second tape).
