CLAYTON@xrt.upenn.EDU.UUCP (06/12/87)
Information From TSO Financial - The Saga Continues... Chapter 4 - June 11, 1987 In response to a question from Geoff Hill of Lakehead University about saving space on a cluster common system disk, Frank J. Nagy suggests that the dump file can be placed in SYS$COMMON for access by multiple systems. We have done the same thing without problems, but have been under the following restriction as told to me by my local DEC guru, and a trip through the internals and fiche. The size of the dump file that is shared must be four (4) blocks larger then the largest memory size of any machine sharing the same dump file. The reason for the four blocks is three are used for ERRLOG buffers, machine state, etc. The fourth is reserved for future use. Understanding the above constraint, the concept of sharing dump files is a valid decision and one that saves CONSIDERABLE space. On one of our cluster system disks, I have five (5) roots which support one 44MB, two 40MB, one 32MB and one 20MB system. Only three roots are active, USUALLY. The others are for failover of systems in the event of CPU crashes. I still have between 50 and 100K blocks left. It should be noted, that on ALL systems, the page and swap files are 10K each on the system disk, and that 100K files for each are on other spindles. Also, the system disk is all most a 'static' volume with the only exception being the target of output for spooled devices when VMS queues are not used, as well as a couple of other areas such as the SYSUAF and JBCSYSQUE files. My definition of 'static' is a volume that does NOT go through the considerable file creation/deletion cycle of a data/user disk. Paul D. Clayton - Manager Of Systems TSO Financial - Horsham, Pa. USA Address - CLAYTON%XRT@CIS.UPENN.EDU