I might be faster to open an SR/PMR with IBM XCF. That way IBM can determine if there is a bug or configuration issue for you
Lizette > -----Original Message----- > From: IBM Mainframe Discussion List [mailto:[email protected]] On > Behalf Of Tracy Adams > Sent: Thursday, October 27, 2016 8:27 AM > To: [email protected] > Subject: XCFAS high CPU after z13s upgrade > > Hi there all, > > Pardon my lack of experience with XCF but I lived in a UNI shell for years now > I am enlightened to the world of XCF. I am wondering what could be wrong here > and looking for your suggestions on where to go... > > Back ground: we were running two 196's using escon CTC connections with a 14 > year old policy using two primary structures (small and large). One dedicated > coupling facility and one LPAR. The shop keeps all the smf data but rarely > runs Monitor III so I am in the process of looking for old type 74 records to > get some comparison data. > > Upgrade: installed two z13s last weekend with ISA links. The VAR rewrote > the CFRM policy to use six structures (1 & 2 for small, 3 & 4 for medium, 5 & > 6 large). The dedicated coupling facility (on the F03 box) is defined for > "thin dispatch". The VAR that installed the z13's recommended this coupling > definition based on the faster ISA links and the use of thin dispatch. > > The net result is that lpar MVSA's address space XCFAS, running on the F03, is > now consuming over the 15 minute interval 10%+ cpu utilization every interval > all day long. Using SYSVIEW and looking at the PLEXGRPS the IORt is pretty > consistent in the 1000 total. The CPU for MVSA's XCFAS though is very jumpy, > sub 1% for a couple of whacks of the enter key, 30-50% for a couple of whacks > of the enter key. The monitor III display shows 86% of the xcf calls going > through small structure 1, 13% of the xcf calls going through structure 3, and > 1% going through structure 5. There is a little that hits structure 2 but it > is insignificant. XCFAS on MVSC, the M03 box running the couple lpar, is < 1% > CPU. Prior to the upgrade XCFAS on both systems averaged < 1% for most 15 > minute intervals and peaked at 3% maybe once or twice a day for an interval. > > The bottom line is the MVSA image does two things for us, IDMS and batch. > MVSC is CICS and DB2. The IDMS/CICS relationship is the majority of the XCF > usage. Batch is getting WLM CPU delays due to high cpu from SYSTEM (xcfas) > even though we added 20% more MSUs in the upgrade. > > I guess the real bottom line question though is why did my combined XCFAS cpu > consumption go from an average of < 1 % on each MVS lpar to 10%+ on one and > stayed the same on the other? > > Thanks in advanced for giving this post some time and energy :) > > Tracy > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
