On 6 Jan 2010 22:19:29 -0800, in bit.listserv.ibm-main you wrote: >Back in OS/390 R5 days I was presented a standalone dump where the exact >same problem (a syntax error in bpxprm on a test system that nobody cared >about but that *was* part of a productive parallel sysplex) had brought down >a productive system due to the fact that the message prompting for the new >member is a synchronous message (so XCF signalling won't work anymore). > >At that time I investigated how that message asking for the new bpxprm was >issued. It was NOT using ASA services back then! Oh, and you can blame the >syntax checker for bpxprm members on me, too! On the grounds that each and >every parmlib member that can stop an IPL should have a way to be checked >before IPL. IBM allowed me to take a FIN apar for that. > >The first time this occured here (years later) I asked everyone changing that >parmlib member to use the syntax checker. The second time it occured, they >got a bashing from me. But from those two occurances I know that the actual >error message (which I believe is now using the ASA services) is issued >hardcopy only, so no amount of HMC config will get you those messages >(unless you have the HMC configured to receive the hardcopy message set - >in which case Health checker will scream). All you'll see on the HMC is the >prompting message. It usually takes either an sadump or backup members to >come up and *see* what the error was. And since this is still NIP, you won't >see those messages in operlog, either, until the prompt is replied to and >NIP/MSI continued. > >As for 'seeing' these hardcopy only messages: I believe that the design idea >was that since all bpx statements are usually 'echoed' in hardcopy log, it is >unnecessary to write out an extra message only stating the error. And who >wants to see all bpx statements (and that they're correct) in the 'normal' >cases? Unfortunately that means that no provision was made to just repeat >the syntax error message which would enormously help. (Same is true for all >those IPL-halting NIP messages like errors in lpalst or progxx - the system >comes up somewhat, but chances are that you don't get to the point of >actually seeing operlog/syslog.) > >As for how to see these messages on the HMC: Those not in the hardcopy set >can be made to appear if you make sure that no other NIP console can be >found by NIP. The way to do that is either close the emulation that has the >NIP console, or if that is impossible due to the location, configure the >channel >path to that (NIP/MCS) console offline from the HMC. That will get you the >console of last resort - the HMC. As far as I am aware, you cannot see NIP >messages on more than one console at a time. > >I have no idea if having the HMC configured to receive the hardcopy set (via >consolexx - which isn't read yet at this time) will actually get you these >hardcopy-only NIP messages.
Maybe I am off base but if a message is about something that can stop an IPL, then it should be viewable on the console where the IPL is taking place. If it isn't and can't be found when IPL quits, I believe that should be a sev 1 APAR with no FIN. As Barbara relates, this can cause a production outage. > >Best regards, Barbara > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
