On Mon, 30 Dec 2024 16:37:12 -0800, Ed Jaffe <[email protected]>
wrote:
>It's good to hear from you brutha. SHARE hasn't killed you yet... ;-)
:)
>Originally, my HZS logstream was variable size i.e., INIT(9M) and
>SIZE(12M) so I changed both values to 9M. No difference in behavior.
>Since our IPL on Sunday, we've seen the IXC534I for OPERLOG only four
>times while it appeared 38 times for Health Checker over that same time
>frame.
I wouldn't expect it to make a difference. In your original post, you had
SIZE(12M) and INITSIZE(9M). The messages indicated HZS_HEALTHCHECKLOG was
already at 12M (as high as it would go). All it can do in this instance is try
to balance the element/entry ratio to better utilize the space it has, yet that
also seems to be set as optimal as it can within the space allotted.
BTW, the Health Checker User's Guide shows the definition with only an SIZE,
yet the CF Sizer yields both a SIZE and INITSIZE. The inconsistency is a
little frustrating. I'm inclined to go with the CF Sizer recommendation. More
below.
>I set the following SLIP to catch the IXLALTER call in the act and examined
>the dump in IPCS:
>
>SLIP
>SET,IF,ID=EEJ1,NUCMOD=IXLCSALT,DATA=(1R?+8,EQ,C8E9E26D),A=SVCD,SDATA=(PSA,NUC,RGN,CSA,SQA,LPA,TRT),END
>
>I found that Health Checker was writing to its logstream (BALR/BAKR to
>IXGW1TSK) and that led to a PC to XCF (IXLC1ALT) to alter the structure.
>LOGGER simply decided it was time to ALTER the structure as a result of
>the current IXGWRITE request.
A thorough analysis as usual... Basically, Logger asks for more, and CFRM/XES
says, "I've given ye all I can, cap'n, I canno' give ye no more." (Apologies to
the late James Doohan.) Lather, rinse, repeat...
>It appears my assumption about which logstream is used more heavily is
>just flat-out wrong. Both structures are roughly the same size. Health
>Checker must be writing ~10 times as much data to its logstream as MVS
>is writing to OPERLOG. It's the only explanation.
Try using SAMPLIB(IXGRPT1) to process a representative set of SMF type 88
records, and you'll know for sure. You can use the data to feed into CF Sizer.
It may just be time to redefine the structure.
>I did notice one difference between the two structure definitions.
>HZS_HEALTHCHKLOG defaults to FULLTHRESHOLD(80) whereas OPERLOG
>explicitly specifies FULLTHRESHOLD(0). I don't know how much that
>matters when there is no automatic altering going on in either case.
If I understand the book correctly, for this issue, it doesn't. Without
AUTOALT, it's only a monitoring setting, but one that Logger or Logger users
might be able to exploit to trigger alter requests ("manual" alter).
>STRUCTURE NAME(OPERLOG) /* OPERATIONS LOG */
> SIZE(8M) /* */
> PREFLIST(CF01,CF02)
> FULLTHRESHOLD(0)
>
>STRUCTURE NAME(HZS_HEALTHCHKLOG) /* HEALTH CHECKER */
> SIZE(9M)
> INITSIZE(9M)
> PREFLIST(CF01,CF02)
>
>Thanks and Happy New Year!
You're most welcome, any time, and Happy New Year to you as well, my friend.
Art
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
