Investigating temporary storage waits
If a user task is waiting on a resource type that starts with TS, showing that it is for temporary storage, the following information describes possible reasons for the wait.
About this task
Procedure
- Resource
type TSAUX indicates that a task is waiting because it has made
an unconditional request for temporary storage, and the request cannot
be met because insufficient auxiliary storage is available. Take the
following actions:
- Get a system dump and format it using the formatting keyword TS to show the temporary storage control blocks, as well as SM and KE as the information for these components could be useful.
- Analyze the dump using the help given in Is temporary storage close to being exhausted? and Is fragmentation of unallocated storage causing the WRITEQ TS request to fail?.
- The task has issued a request requiring too large a piece of temporary storage.
- The task has issued a request requiring a reasonable amount of temporary storage, but there is too little available.
The task has issued an EXEC CICS WRITEQ TS command, without specifying NOSUSPEND and without any code to handle the NOSPACE condition. If SPURGE(YES) is defined for the task on the CEDA DEFINE TRANSACTION command, and a deadlock timeout interval other than 0 has been specified, the task is purged when that time expires. Otherwise, it is not purged, and is liable to be suspended indefinitely.
A task that makes a conditional temporary storage WRITEQ TS request (NOSUSPEND specified) is not suspended if the request cannot be met. Instead, if the required auxiliary storage is not available, an exception response is returned to it. There might still be a suspension for another reason - for example, the temporary storage program itself might become suspended after issuing a GETMAIN, if CICS® went short on storage.
- Resource type TSBUFFER indicates that the task that is waiting has issued an auxiliary temporary storage request, but the buffers are all in use. If you find that tasks are often made to wait on this resource, consider increasing the number of auxiliary temporary storage buffers (system initialization parameter TS).
- Resource type TSEXTEND indicates that the waiting task has issued a request to extend the auxiliary temporary storage data set, but some other task has already made the same request. The wait does not extend beyond the time taken for the extend operation to complete. If you have a task that is waiting for a long time on this resource, check if there is a hardware fault or a problem with VSAM.
- Resource type TSIO indicates that the task is being made to wait while physical I/O takes place during an auxiliary temporary storage read or write. If there is an extended wait on this resource, check if there is a hardware fault or a problem with VSAM.
- Resource type TSMAINLM indicates that the task that is waiting issued a main temporary storage request, but the request cannot be met because insufficient storage is available. If you find that tasks are often made to wait on this resource, check for large temporary storage queues that could be deleted, or consider increasing the limit for the storage that is available for main temporary storage queues to use (system initialization parameter TSMAINLIMIT).
- Resource type TSPOOL indicates that the maximum number of concurrent requests (10) for a temporary storage pool in the coupling facility has been reached. The task resumes when one of the requests completes.
- Resource type TSQUEUE indicates
that the waiting task has issued a request against a temporary storage
queue that is already in use by another task.
The latter task has the lock on the queue. The length of time that a task has the lock on a temporary storage queue depends on whether or not the queue is recoverable. If the queue is recoverable, the task has the lock until the logical unit of work is complete. If it is not recoverable, the task has the lock for the duration of the temporary storage request only.
If tasks in your system are frequently made to wait on temporary storage queues, consider the following:
- Are tasks that are performing operations on the same temporary storage queue intended to do so, or is the ID of the queue unintentionally not unique?
- Is it possible to create more temporary storage queues to reduce the contention between tasks?
- If the queue in question is recoverable, is it possible to make tasks relinquish control of it more quickly? Consider reducing the size of UOWs, or making conversational tasks pseudoconversational.
- Resource type TSSHARED indicates that a shared temporary storage request is being processed asynchronously. The task is resumed when the request is complete. The logic that determines whether a request is processed synchronously or asynchronously is outside CICS control.
- Resource type TSSTRING indicates that the task is waiting for an auxiliary temporary storage VSAM string. If you find that tasks frequently wait on this resource, consider increasing the number of temporary storage strings (system initialization parameter TS).
- Resource type TSWBUFFR indicates that the waiting task has issued an auxiliary temporary storage request, but the write buffers are all in use. You have no control over how temporary storage allocates read buffers and write buffers from the buffer pool, but if you find that tasks are often made to wait on this resource, increasing the number of auxiliary temporary storage buffers (system initialization parameter TS) should help solve the problem.
- Is temporary storage close to being exhausted?
It could be that your task has made a reasonable request for temporary storage, but the amount of unallocated space is close to exhaustion. - Is fragmentation of unallocated storage causing the WRITEQ TS request to fail?
You can tell whether fragmentation of unallocated temporary storage is causing the WRITEQ TS request to fail by looking at the temporary storage summary in the dump.
dfhs11e.html |
Last updated: Monday, 5 June 2017