Investigating storage waits

If a task is waiting for a long time on any of the resource types CDSA, RDSA, SDSA, UDSA, ECDSA, ERDSA, ESDSA, ETDSA, EUDSA, GCDSA, GUDSA, or GSDSA, you can use a CICS® system dump to investigate further.

About this task

Waits on these resources occur when tasks make unconditional storage requests (SUSPEND=YES) that cannot be satisfied. For storage requests below the 16 MB line, waits can be on the CDSA, RDSA, SDSA, or UDSA. For storage requests above 16 MB but below 2 GB, waits can be on the ECDSA, ERDSA, ESDSA, ETDSA, or EUDSA. For storage requests above the bar, waits can be on the GCDSA, GUDSA, or GSDSA.

If tasks make conditional storage requests (SUSPEND=NO), tasks are not suspended on these resources. If the request cannot be satisfied, an exception response is returned.

CICS automatically attempts to relieve storage when it is under stress, for example by releasing storage occupied by programs whose current use count is 0. Also, if a task waits for storage longer than the deadlock timeout parameter specified in the installed transaction definition, the task might be automatically purged. Certain conditions prevent purging of a task, for example, a deadlock timeout value of 0, or a specification of SPURGE(NO).

The most likely reasons for extended waits on storage requests are as follows:
  • The task issued an unconditional GETMAIN request for an unreasonably large amount of storage.
  • The task issued an unconditional GETMAIN request for a reasonable amount of storage, but the system has too little available. The system might be approaching a short-on-storage (SOS) condition, or the storage might be too fragmented to satisfy the request.

Procedure

  1. Produce a CICS system dump, and format it using the formatting keyword SM.
  2. Interpret the dump to investigate the problem.
    1. Was the request for too much storage? Look in the SM suspend queue summary.

      This summary includes the number of bytes requested by every task that the storage manager has suspended. You can see whether any task made a GETMAIN request for an unreasonably large amount of storage.

      For example, the following is the dump output in the SM suspend queue summary when task 41 requests 10,000,000 bytes:
       ==SM: Suspend queue summary
       
          KE Task  Tran #  Susptok  Subpool  DSA       Request
       
          053E5400 0000041 04080011 U0000041 EUDSA    10000016
      If the suspended task made a reasonable GETMAIN request, investigate whether the system is approaching an SOS condition.
    2. Is the storage close to being exhausted? Look at the DSA summary in the formatted dump.

      This summary shows the current free space in each DSA, both in bytes and as a percentage of the total storage. It also shows the size of the largest free area, that is, the biggest piece of contiguous storage. (In this context, contiguous storage means storage that is not fragmented by other records. It is accepted that records too large to fit in a single CI can be split across two or more CIs that are not necessarily contiguous.)

      If the largest free area is smaller than the requested storage, this is probably the reason why the task cannot obtain its requested storage.

      If the amount of free space is unexpectedly small, look at the task subpool summary. If a task has made an unusually large number of GETMAIN requests, this might indicate that the task is looping. A looping task might be issuing GETMAIN requests repetitively, each for a reasonable amount of storage, but collectively for a very large amount. If you find evidence for a looping task, see Dealing with loops for more information.

      If the task made a reasonable storage request and the system seems to have sufficient free storage, investigate whether fragmentation of free storage is causing the GETMAIN request to fail.
    3. Is fragmentation of free storage causing the GETMAIN request to fail? Look at the DSA summary in the formatted dump. If the summary shows that the current free space is significantly greater than the largest free area, it is likely that the DSA has become fragmented.


dfhs11d.html | Timestamp icon Last updated: Monday, 5 June 2017