[309] in Athena_Backup_System
Minutes of 19 Sept Meeting
daemon@ATHENA.MIT.EDU (Bill Cattey On the Road)
Thu Sep 19 14:20:46 1996
Date: Thu, 19 Sep 96 13:55:01 EDT
From: wdc@MIT.EDU (Bill Cattey On the Road)
To: athena-backup@MIT.EDU
Bill will be attending only the beginning of the ABS meetings until the
current review of the design is done. This is because he's opting out
of having the detailed knowledge of the system, and focusing on other
aspects of managing the project.
The ABS meeting format is now:
First 15 Minutes: status reports and review of open issues.
The Rest: The detailed design review.
This week Miki and Diane came up with a change to how dumpsets are
managed that simplifies the code and makes the system easier to
understand:
A dumpset is a set of specifications of server, partition, and volume
where volume can be wild carded. The enumeration of volumes happens
at dump time (just like AFS does it.) The specific volumes that ended
up on a particular tape is recorded in the database as a consequence of
the dump status report.
The order of events in a partition or server restore is this:
1. The vldb is queried to find out what volumes are known to be part of
the partition or server at the present time.
2. The database is updated with the correct location of the volumes.
(Volumes might have moved since the dump was taken.)
3. Then the database is queried for all tapes relevant to the volumes to
be restored.
4. The server or partition restore defaults to restoring the most
recently dumped copy of the volume.
5. There is the possibility that a volume was new enough not to have a
backup. Volumes that are requested for restore that are not on any tape
are reported. (A complete log of everything that does and does not get
restored is reported.)
We also discussed that a "tell me what you were going to restore, but
don't perform the restore" would be useful. Or "Set up the restore"
(i.e. do the update in #2 above) followed by "OK. Do it.".
----
No news on Crippled Mode.
The large cell test was suspended while Miki and Diane worked out the
simplification of the Dumpset Management. It will resume soon. (Note:
Oracle Web HAS been moved off metamucil for the security of the test.)
Diane says that now that we have the secure web prototype, we should
investigate ABS user interfaces via the web. Clicking in a web form is a
LOT cleaner than Tcl/Tk.
-----
Review of section 7 is suspended until Diane rewrites the document to
account for the design change on managing dumpsets.
Then Bill left and the review resumed.
-wdc
