Re: Corrupted IPL Record

[O'Brien, Dennis L]

Richard, Yes, ending VM:Secure before reallocating the object directory disk would be sufficient.  I'd do the same on a system with DIRMAINT.  Even if DIRMAINT doesn't cache the allocation map, I wouldn't want it updating the active cylinders while I was reallocating them.                                                                         Dennis           There are 10 kinds of people in the world; those that understand binary and those that don't.   From: The IBM z/VM Operating System [mailto:[EMAIL PROTECTED] On Behalf Of Schuh, Richard Sent: Monday, October 30, 2006 16:19 To: IBMVM@LISTSERV.UARK.EDU Subject: Re: [IBMVM] Corrupted IPL Record No VM:Secure. The problem occurred on a small, special purpose machine that is accessed only via Secure TN3270. There are only 4 terminal addresses defined and they are all logged on in a secure room. There are no network connections except for an NJE link to our main VM system. It is used for submitting jobs and sending files to the appropriate z/OS systems. It is one-way communication. There isn’t any need for a heavyweight ESM.   This was a new error to me. The normal sequence of an IPL is:   Record 0/0/1 is read in to location 0. It contains the IPL PSW, IPL CCW1 and IPL CCW2. A TIC to IPL CCW1 is done. CCW1 reads the first record of the initialization program. CCW2 is executed. it either reads the rest of the IPL program or TICs to a location in the record just read to load the rest of the init pgm. When the channel program ends successfully, the IPL PSW is loaded and the init pgm loads and starts the O/S.      The allocation map is not used until after CP is started. We were failing when step 4 was supposed to occur. The first record of the init pgm was the one that was corrupted. Record 4, where the allocation map lives, was not touched.   Is there something in the system that arbitrarily rewrites these IPL records? How about record 3, the volume label? The pseudo VTOC (records 5 and 6)?   As for the VM:Secure thing, I find that simply documenting how the product can screw up a system is not only an unacceptable answer, it borders on repugnance. A strategic product that should always be working should never violate the integrity of the system. There is no justification for it. The changed allocation on the disk must be respected. Reallocating a disk is a normal maintenance activity. If the documentation change is the answer, someone needs to update the documentation for CPFMTXA/ICKDSF with a very stern warning about the potential for disaster if VM:Secure is running when a disk is being reallocated.     Would I be presumptuous in thinking that ending VM:Secure before reallocating the disk would be a sufficient precaution? Even that would be a problem for us. The Rules Facility is heavily used in our environment.   Regards, Richard Schuh   -----Original Message----- From: The IBM z/VM Operating System [mailto:[EMAIL PROTECTED]On Behalf Of Mike Walter Sent: Monday, October 30, 2006 1:02 PM To: IBMVM@LISTSERV.UARK.EDU Subject: Re: Corrupted IPL Record   Were you running VM:Secure on that system?  Are the DRCT cylinders on that IPL DASD?  If so, this may help. When VM:Secure starts up, it reads the whole allocation bit map of the DASD with the source directory minidisk (usually: VMSECURE 01B0).  Each time VM:Secure rebuilds the object directory cylinders (msg: "VMXRXB0740I The dynamic REBUILD has completed.  Directory maintenance activity will now resume.") it completely re-writes ALL of the allocation bitmap (as it was when VM:Secure came up) from its cached copy, except for updates to the bits in the DRCT cyls. That bit us (excuse the pun) a couple Sunday IPLs in a row when the newly expanded PARM disk kept getting changed back to its old size.  The back end of the newly updated SYSTEM CONFIG happened to have 4K blocks allocated on the new cylinders.  When VM:Secure re-wrote the PARM allocation map size/location (begin location did not change, just the end), it chopped off the last half or so of the SYSTEM CONFIG.  CP happily came up without error because the truncation just happened to be between SYSTEM CONFIG statements.  To diagnose it, I wrapped SYSTEM CONFIG with confirmatory messages issued as it runs:    Say "Beginning: 'SYSTEM CONFIG' from MAINT's CF1 disk..."      TOLERATE_CONFIG_ERRORS NO                                     ... rest of statements...    Say "Completed: 'SYSTEM CONFIG' from MAINT's CF1 disk." Be careful of "TOLERATE_CONFIG_ERRORS NO".  Don't just drop it in without trying it live.  There are non-syntactical errors which will pass CPSYNTAX (everyone DOES run CPSYNTAX **EVERY TIME** after changing SYSTEM CONFIG, right?), but will cause an IPL error.  In our case, the missing 'Completed' statement confirmed the suspicion.  And... explained why the system came up so half-configured (missing the last half or so of SYSTEM CONFIG). After reporting it to CA, they said they would update the doc, showing how VM:Secure can cause these sorts of problems. By chance, I had a conversation with the CPSYNTAX developer about an open PMR just last week.  I suggested some type of new statement pairs which, if present, must BOTH be present as the first and last non-comment records in SYSTEM CONFIG (and perhaps IMBED files) to diagnose just this sort of error. Mike Walter                                                         Hewitt Associates                                                   Any opinions expressed herein are mine alone and do not necessarily represent the opinions or policies of Hewitt Associates.             "Schuh, Richard" <[EMAIL PROTECTED]> Sent by: "The IBM z/VM Operating System" <IBMVM@LISTSERV.UARK.EDU> 10/30/2006 02:32 PM Please respond to "The IBM z/VM Operating System" <IBMVM@LISTSERV.UARK.EDU> To IBMVM@LISTSERV.UARK.EDU cc   Subject Corrupted IPL Record   If this shows up twice, I apologize. I first sent it 2 hours ago and it hasn't hit the archives as yet. Over the weekend, we upgraded our final system to 5.2. As a part of the migration, we changed old PARM extents to PERM and allocated new PARM extents. When we tried to ipl the system, there were errors that led me to the conclusion that the IPL program had been corrupted. Using DDR on another system, I observed that record 0 0 2 had been completely wiped out. There were a few scattered bits that were not 0 in the first 100-200 bytes, nothing that was any kind of pattern, and the rest of the record was all 0. Records 1, and 3-6 were all as they should have been. I had to  run SALIPL to fix the IPL program. I am not blaming CPFMTXA ALLOCATE because I think it highly unlikely that it was the cause. I suspect that something else happened between the previous IPL and yesterday's failed attempt. Has anyone else seen this kind of corruption or are we, once again, unique? Regards, Richard Schuh The information contained in this e-mail and any accompanying documents may contain information that is confidential or otherwise protected from disclosure. If you are not the intended recipient of this message, or if this message has been addressed to you in error, please immediately alert the sender by reply e-mail and then delete this message, including any attachments. Any dissemination, distribution or other use of the contents of this message by anyone other than the intended recipient is strictly prohibited.