[200] in Athena_Backup_System
Re: "cleanly"
daemon@ATHENA.MIT.EDU (dkk@MIT.EDU)
Fri Feb 23 11:03:43 1996
From: dkk@MIT.EDU
Date: Fri, 23 Feb 1996 11:03:18 -0500
To: delgado@MIT.EDU
Cc: athena-backup@MIT.EDU
In-Reply-To: <9602231400.AA15985@green-acres.MIT.EDU> (message from Diane Delgado on Fri, 23 Feb 1996 09:00:53 EST)
dkk>- It is *NOT* acceptable that the DB become available for use
dkk> (e.g. after a reboot) while its consistency is still in doubt
dkk> (i.e. before some sort of consistency checker is run).
delgado> As usual, I'd like more information on what the "consistency
delgado> checker" is supposed to be.
There isn't necessarily supposed to be one. The requirement is just
that the DB not go into service if it is corrupt, thereby magnifying
the corruption. If an administrator does something boneheaded, we
expect trouble. But the master should not come up ready to run
backups if the database is screwed up.
The consistency required is both in Oracle and in the ABS data. It is
sufficient for Oracle to keep the ABS data consistent by keeping
operations atomic by committing them only at appropriate points, so a
half-done transaction is lost in a crash. If, for some reason, the
configuration of the Oracle DB cannot assure consistency, then we'd
need some external assurance.
delgado> By the way, the DB's availability after such situations is
delgado> strictly up to the adiminstrators (this is a policy issue)
delgado> and shouldn't be listed as a requirement.
It is not a policy issue if the DB becomes available on its own,
without human intervention, and without concern for local
configuration details.
delgado> I certainly wouldn't object to restating this requirement as:
delgado> "There must be a means available to verify/ensure that the DB
delgado> is in a consistent state while it is closed/offline ".
That would be one way to meet the requirement, as I understand it and
wrote it, assuming the DB does not become open/online automatically
(e.g. upon reboot).
Was that clear?
