PH66956: IMPROVE RESILIENCE OF KEEPALIVE REQUESTS WHEN A PROCESS HAS NO IDLE THREADS

APAR status

• Closed as program error.

Error description

• Under some load patterns, clients may see a connection closed
  shortly after sending a keepalive request.
  

Local fix

Problem summary

• ****************************************************************
  * USERS AFFECTED:  All users of IBM HTTP Server on Linux or    *
  *                  zOS                                         *
  ****************************************************************
  * PROBLEM DESCRIPTION: Under some load patterns, clients may   *
  *                      see a connection closed                 *
  *                      shortly after sending a keepalive       *
  *                      request.                                *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  Consider a configuration with the following MPM settings:
  MinSpareThreads  25
  ThreadsPerChild  25
  MaxSpareThreads 250
  MaxClients 2500
  Normally we only have problems as we approach 2500 clients /
  threads. However, if by some sequence of load we end up with
  any
  single process using all of its threads, some bad behavior can
  result even when there is excess capacity in other processes.
  When a process is at 25/25 busy threads and a keepalive
  connection has new activity, IHS will respond to that activity
  by closing the connection. Unfortunately, there is nearly no
  logging of this condition.
  Similarly, but in a much more unlikely path, all keepalive
  connections may be closed when the server is in the above
  state.
  

Problem conclusion

• The code was updated to provide a better default behavior and
  logging in this area.
  
  1. By default, when the server is in the state described above,
  activity on a keepalive connection is queued for the
  next available worker rather than closed.
  
  When this queueing occurs, two informational messages are
  logged: IHS00035/IHS00036
  
  Connections queued in this way are marked to close
  gracefully at the end of the upcoming request/response.
  
  To restore the previous behavior, set global directive
  'IHSKeepaliveCloseOnActivity ON'. In this case, a warning
  level message is issued with identifer IHS00037
  after the connection is closed.
  
  2. By default, in the state described above, all keepalive
  connections are instead marked for closure after the next
  request and response.
  
  The previous behavior can be restored with
  'IHSKeepaliveCloseAll ON' in which case a warning message
  with identifier IHS00038 will be issued.
  
  3. The server periodically checks for the state described above
  and marks keepalive connections as being closeable after the
  next response.  Informational message with identifier IHS00041
  is issued when this occurs.
  
  The fix for this APAR is targeted for inclusion in
  IBM HTTP Server fix packs 9.0.5.25 For more information, see
  'Recommended Updates for WebSphere Application Server':
  https://www.ibm.com/support/pages/node/715553
  

Temporary fix

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

• APAR is sysrouted TO one or more of the following:

Fix information

• Fixed component name

  IBM HTTP SERVER

• Fixed component ID

  5724J0801

Applicable component levels