I praise the development staff for providing an ADSM solution for
automatic scheduling of the Macintosh platform. However, I have some
issues with the implementation. First, the memory requirements for the
application have always been traditionally lower than the real
requirements needed for the application to function properly.
The scheduler uses 4 meg + to run successfully with a new node backing
up from scratch without generating 'Memory Running Low' errors in the
middle of a backup. It is this 'fudging' on the memory numbers that
makes the appearance from time to time of the ADSM scheduler running in
the background on the client to be exactly opposite to the the impact
the documentation states. And although it only uses a fraction of this
total when it is polling the server for a schedule, it still has to be
allocated just in case there is actually work to be performed. We
currently have nodes poll the ADSM server every 12 hours to minimize the
traffic on our limited network. Nevertheless, the appearance of a 4 meg
+ program counting down for 12 hours becomes an memory allocation
problem on some smaller machines. Short of having users form the habit
of terminating the program, any suggestions of what we can do on the
server or client side would be appreciated. Our dream solution would be
for the client scheduler to poll the server for a DATE as well as time
and finding the current date is not equal to the backup window date, to
quit and thereafter poll the server in the next polling period.
Another solution that encounters a problem:
Before the introduction of the Schedule Daemon, we were able to use
another application called CRON to kick start the node's scheduler. The
advantage of CRON was that it only used 96K and it was totally invisible
to the user. You could also tailor it to execute the client scheduler
within minutes of that node's backup window, cutting down the time the
node needed to poll the server. However, although CRON still works for
this purpose, the logging of the node connection and its backup is no
longer sensed by the ADSM Scheduler on the SERVER. The server now flags
the node as 'Missing its backup'. This is puzzling because the
activity log records the node's connection at the proper time for the
backup window and the client scheduler executes the backup without
problems. I know ADSM is under no obligation to support third party
solutions, but shouldn't the ADSM Scheduler on the server recognize that
a node has connected during its backup window? FYI: We are running
server version 3.1.0.3 on an AIX system.
******************************************
Renee Davis
Central Computing/Telecommunication Services
Program Analysis Specialist
RDavis AT UH DOT EDU
******************************************
|
