[77] in Athena_Backup_System

home help back first fref pref prev next nref lref last post

meeting minutes 2/9/95

daemon@ATHENA.MIT.EDU (Diane Delgado)
Fri Feb 10 10:55:26 1995

To: athena-backup@MIT.EDU
Date: Fri, 10 Feb 1995 10:55:08 EST
From: Diane Delgado <delgado@MIT.EDU>


Here's the summary of Thurday's meeting.  Jonathon and Dave, please
feel free to correct any mistakes.  (It's long, but we did talk 
about many important issues).)

attendees: Jonathon, Dave, Diane


- The Tape format for the ABS was discussed.  Dave presented
  some interesting factiods regarding tapes and eof markers.
  The result is we need to try out a few things to determine
  what the features are for the drives we are using, and then
  modify the proposed tape format, if necessary.

-We began discussing the operational requirements for the
 ASD Database Services support.

   - 7x24 support from ASD DS is a requirement from DCNS operations.
     There were no objections to this.

   - The recovery time after any database failure should be limited
     to 4 hours.  This is also a DCNS operations requirement.
     There were no objections to this.

   - The allowable downtime for the ABS database was discussed.
     We felt that we could allow some downtime on a weekly basis.
     No one was certain whether daily downtime was allowable.
     We also do not yet have a clear idea of what is  the maximum
     allowable downtime per day or per week.


   - Jeff requies that the Database Backups be in ascii format.
     We will investigate Dave's idea of performing a binary
     dump of the database, and then generating an ascii dump
     from that binary dump.  This has a few advatages:
     (a) binary dumps are much faster, thus reducing the time
     the backup interferes with the regular DB activity.
     (b) staff can use the binary backup for disaster recovery
     (which is faster) and revert only to the ascii if using the
      binary dump fails.  We need to look into performance issues
      to ensure that this extra step does not place an excessive load
      on the DB machine.


   - The issue of who should have DBA access was discussed.
     (This issue is not yet fully resolved.)  Jeff requires that
     operations have full DBA access to the Oracle Database; such
     DBA access would be restricted to only those staff who currently
     have DBA access to the Moira Database.    Jeff is against requiring
     that those who have access recieve proper training before access
     is granted.

     Within the team there are varying levels of agreement with this
     proposal and some concerns:

     - the cost of the service agreement with ASD DS might be more
       expensive since they would have to account for possibly having
       extra work due to having persons without the proper training
       accessing the database components and also that changes may
       be made without their knowledge which could adversly affect
       system behavior.

     -Who did it?  With more people having access, how to ensure
      accountability for actions must be considered.

     -With people in different IS groups managing the database
      how do we ensure everyone properly communicates any 
      changes which are made to everyone who needs to know.
      This is essential in quick problem resolution.

     -The fact that DCNS has demonstated that it does not have
      the skill set to perform DBMS management raises the question
      of why anyone in DCNS should have full privileges, especially
      without proper training.

     Everyone was in agreement that at least read-only access to the DB
     along with shutdown and startup privileges was acceptable.
     The issue of privileges beyond this is still unresolved.

     It was also more or less agreed that if we went the route of
     requiring DBA training to get the DBA privilege, that something
     on the order of a 2-day course would be acceptable.
   

   -(can't remeber if we covered this, but I'll add it so that it's
     written somethere)
     
     Jeff wants to ensure that ADS DS gives our Database equal priority
     with the other databases they are managing.  We need to address
     his concern that during a major disruption in service, wherein
     multiple databases are requiring service, that ours is not left
     until last.  The message here is that the ABS is a very important
     service to us, and can become critical during disaster recovery.
     In the event of a disaster, we need to have it available as
     soon as possible.


home help back first fref pref prev next nref lref last post