PH27505: WLM CAN GET IN A LOOP WHEN RECEIVING COMPRESSED DATA RUNNING ON Z15 HARDWARE

APAR status

• Closed as program error.

Error description

• When WLM is decompressing data received it can get in a loop,
  the following stack can be seen in javacore:
  
  at java/util/Arrays.copyOf(Arrays.java:3580(Compiled Code))
  at
  java/io/ByteArrayOutputStream.grow(ByteArrayOutputStream.java:12
  9(Compiled Code))
  at
  java/io/ByteArrayOutputStream.ensureCapacity(ByteArrayOutputStre
  am.java:104(Compiled Code))
  at
  java/io/ByteArrayOutputStream.write(ByteArrayOutputStream.java:1
  64(Compiled Code))
  (entered lock: java/io/ByteArrayOutputStream@0x000000002A454E50,
  entry count: 1)
  at
  com/ibm/ws/cluster/Compressor.decompress(Compressor.java:219(Com
  piled Code))
  (entered lock: java/util/zip/Inflater@0x0000000019BA5000, entry
  count: 1)
  

Local fix

• N/A
  

Problem summary

• ****************************************************************
  * USERS AFFECTED:  Users of IBM WebSphere Application Server   *
  *                  Network Deployment using Linux on z15       *
  *                  hardware                                    *
  *                  with hardware compression enabled           *
  ****************************************************************
  * PROBLEM DESCRIPTION: An apparent hang occurs in the server,  *
  *                      with high CPU consumption and steady    *
  *                      memory growth                           *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  A defect (IJ26478) exists in Linux on zSeries Java levels prior
  to 8.0.6.16 wherein, when hardware compression is enabled on z15
  systems, the compressed data is invalid in a way that can
  prevent decompression processes from detecting that the
  decompression has actually finished.
  In this specific case, the Work Load Manager component
  compresses cluster description data to send to other processes.
  When that data is decompressed, the failure to read the end of
  the compressed data results in an endless loop that also adds
  bytes to a data structure. In that case, the process will hang
  until eventually running out memory.
  

Problem conclusion

• Mitigation was added to the Work Load Manager component to avoid
  the endless loop. Note that there are other potential issues
  caused by the faulty data compression, and for that reason
  IJ26478 should be applied even if this APAR is present.
  
  The fix for this APAR is targeted for inclusion in fix packs
  8.5.5.19 and 9.0.5.7. 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

  WEBS APP SERV N

• Fixed component ID

  5724H8800

Applicable component levels

• R850 PSY

     UP

• R900 PSY

     UP