SMS Control Dataset Contention

Resolving The Problem

Check the following items when there is contention:

1. Verify the share options. The SCDS is the only CDS that should have
SHROPTNS(2,3). The ACDS and the COMMDS should be SHROPTNS(3,3). If not,
you have a window for contention.

2. When there is a serialization problem, we suggest to verify the
recommendation in the DFSMS Implementing System-Managed Storage manual
sc26-7336, Chp4: Defining GRS Resource Names for SMS Control Data Sets.
The same will apply if you are using MIM. The recommendation is to
convert hardware RESERVES to SYSTEMS scope enqueues. Between systems,
the data sets are serialized by MVS RESERVE/RELEASE. If you allocate a
data set on a device with other data sets protected by RESERVE/RELEASE,
it locks out all activity from other systems to that volume. We
recommend that you use global resource serialization (GRS) to convert
the reserve into a global ENQ.

From the DFSMS Migration manual:
"To avoid contention problems, use global resource
serialization (GRS) to convert hardware reserves on the ACDS
and COMMDS to SYSTEMS scope enqueues . Reserves are issued
with a major name of IGDCDSXS and a minor name of the data
name. You should place resource name IGDCDSXS in the RESERVE
conversion resource name list (RNL) as a generic entry. This
will minimize contention delays and prevent deadlocks which
are sometimes associated with the VARY SMS command."

From experience this section usually means that you received the
following symptoms -

a) IP ANALYZE RESOURCE:
RESOURCE #0014 IS HELD BY:
JOBNAME=SMS ASID=0029 TCB=007D9E88
DATA=SUSPENDED AND NOT DISPATCHABLE

b) SYSLOG Messages:
IEF196I IOS071I 1013,**,SMS, START PENDING
IOS431I DEVICE 1013 RESERVED TO CPU=0C99FF2097,LPAR ID=0C 536
SYSTEM=JC0 MODIFIED BY CPUIDMPF

The IOS431I message is informing us that there is a hardware reserve
held by CPU 0C99FF2097. These numbers break down as follows:
CPU ID breaks down as 2097 (last 4 digits) then serial number x99FF
and LPAR ID=0C. The system holding the long reserve is identified
in the IOS431I message as system JC0. D GRS,DEV=xxxx that would
detail more informaiton on the hardware reserve. However this
typically means that you need the following definition in
SYS1.PARMLIB(GRSRNL00):

RNLDEF RNL(CON) TYPE(GENERIC) QNAME(IGDCDSXS)

3. It is recommended that SMS control data sets "not" be placed on
volumes with high i/o activity, nor should the SMS control data set be
placed on the same volume.

4. IBM recommendation is to issue the GLOBAL commands vs Routing. The
MVS SYSTEM COMMANDS SA22-7627, Chapter 1: Using Commands That Have
Sysplex Scope, presents other commands, that depending of the parameters
coded, they are treated as "scope" commands. Also, this topic explains
the use of a scope command versus ROUTE *ALL

5. Check DINTERVAL and INTERVAL values. How frequently you want the
systems to read for updates/statistics.

6. From the DFSMS Storage Administration Reference:
"GRS serialization only work within the scope of a single Parallel
Sysplex:
An SMS complex consists of systems or system groups that share a
common configuration. A Parallel Sysplex is made up of systems
that share a cross-system coupling facility (XCF); you can run
multiple SMS complexes within a Parallel Sysplex.
IBM strongly recommends that an SMS complex not span sysplexes.
All of the volumes in the SMS complex should be in the same
Parallel Sysplex, since cross-system sharing functions, such as
VSAM record-level sharing (RLS), PDSE sharing, RACF security and
global shared resources (GRS) serialization only work within
the scope of a single Parallel Sysplex. These types of functions
are not supported when the SMS complex extends beyond the
Parallel Sysplex in which they are carried out."

7. OA14084 z/OS 1.5 thru z/OS 1.7 - Contention on SYSZIGDI causing
applications that monitor RMF activity to back up and can also cause UCB
SMS bit getting turned off.

8. OA11917 Hardware resource constraint problem.

If you are using MIM, please check the following items:
1. Upgrade to the latest release and make sure to convert all your SMS
resource definitions from SELECT to ALLSYSTEMS mode.

2. Please review BDC000009595 (Servicelink) at topic 'SOLUTION' for a
detailed description of a set-up problem where SMS starts before MIM.
3. Ensure TDB6027 / QO94415 is applied to resolve a problem where
IEF211I and IGD17204I messages were issued with MII 11.6.
4. Ensure TDB6030 / QO94558 is applied to resolve an abend0c4 rc04 that
can occur with MII 11.6.