[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference iosg::all-in-1_v30

Title:*OLD* ALL-IN-1 (tm) Support Conference
Notice:Closed - See Note 4331.l to move to IOSG::ALL-IN-1
Moderator:IOSG::PYE
Created:Thu Jan 30 1992
Last Modified:Tue Jan 23 1996
Last Successful Update:Fri Jun 06 1997
Number of topics:4343
Total number of notes:18308

1271.0. "Does Backup of PDAF require shutdown?" by KAOFS::D_STREET () Wed Aug 19 1992 16:22

    Hello,
     I have a customer with about 4,000 ALL-IN-1 users who is concerned
    about the validity of their backups. In reading the ALL-IN-1
    Information update (May 1992), they learned of the importance of
    shutting down ALL-IN-1 during backups due to the possible corruption of
    the SDAF. They point out that the PDAF has the same format, and should
    have the same potential problem.
    
     Could I get confirmation on this as the client wants me to fax them
    the indicating the "fact" that they will have to shut down ALL-IN-1
    during the entire backup process. (or set up disk shadowing, or
    something)
    
    					Thanks
    
    					Derek Street.
T.RTitleUserPersonal
Name		DateLines
1271.1No disasterUTRTSC::SCHOLLAERTVivaceObjectTeamWorkLinks for IOSWed Aug 19 1992 17:4514
    Derek,
    
    >the SDAF. They point out that the PDAF has the same format, and should
    >have the same potential problem.
    
    I don't think the FORMAT matters, files might get corrupt when
    data is WRITTEN to them during backup. For a PDAF,
    its not a big problem to recover, because each record 
    has an equivalent in DOCDB. TRU will fix it. 
    
    Regards,
    
    Jan
1271.2PDAF = one user, SDAF = all usersSCOTTC::MARSHALLPearl-white, but slightly shop-soiledWed Aug 19 1992 17:529
The difference between PDAFs and SDAFs is that only one user writes to a PDAF
(their own), whereas all users write to the SDAFs.  Therefore, there is
more likelihodd of the SDAF being written to (and hence possibly corrupted)
while BACKUP is trying to read it, than with a PDAF.

Of course, you'd have thought that BACKUP would have sufficient file locking
to prevent this happening at all...

Scott
1271.3But you told BACKUP to ignore locking ...AIMTEC::PORTER_TTerry Porter, ALL-IN-1 Support, Atlanta CSCWed Aug 19 1992 23:0343
The problem with potential corruption of files is when you use the
/IGNORE=INTERLOCK qualifier and the data in the file gets modified.

The /IGNORE=INTERLOCK qualifier tells BACKUP to copy the data ignoring any
locking that may be going on, hence BACKUP can not guarentee the integrity
of the file.

For performance BACKUP does not do record transfers, but transfers one or more
buckets at once. If the updates the users do to the file do not cause any
change to the structure of the buckets in the file then all you lose is any
changes written to parts of the file BACKUP has already copied. However if an
update to the file causes a bucket split you could get into the situation where 
BACKUP moved a bucket before the split and then later encounters part of the 
same bucket (but after the split). In this case the RMS structure of the
resulting file is not consistant and hence when the file is restored RMS can
not access it properly.

The Info Update artical describes ways to minimise/remove the danger of this
corruption and I would strongly suggest that the customer addopts the solution 
most suited to him described in the artical.

If they are concerned about existing BACKUPs then they can be checked out by
restoring the file to disk and doing an ANALYZE/RMS on the file. If no errors
are found the file is OK, if errors are found then the CSC has tools that
can fix the corruption (with a small chance of some data loss depending exactly
what is corrupted).

This tool is not available outside the CSC, but if the fixing of the file is
critical to the running of the customer's production system (i.e. they need 
to restore the backup to fix a problem but discover that the file on tape is 
corrupt) then we will fix that file under the customer's support contract,
assuming they have one! (in this situation the customer should log a call 
with the CSC in the normal manner).

If fixing the file is not critical then we can do it for a charge (in this case 
contact Wendy Ellis @ALF for detailed information).

Note: The above applies to the US CSC only, I do not know what service the other
CSC provide.

HTH

Terry