LI80623: AVOID CASSANDRA OUT OF DIRECT MEMORY ERROR BY REDUCING THE FETCHSIZE

APAR status

• Closed as program error.

Error description

• During a Dynamic Reregistration and Reconfiguration (DRR) of
  the Gateway, the following error is seen in the
  default_apiconnect-apiconnect-cc-<id> pod logs:
  ERROR [epollEventLoopGroup-2-1] <timestamp> Message.java:629 -
  Unexpected exception during request;
  io.netty.util.internal.OutOfDirectMemoryError:
  failed to allocate 16921 byte(s) of direct memory (used:
  67101489, max: 67108864)
   at
  io.netty.util.internal.PlatformDependent.incrementMemoryCounter
  This error caused a failure in the DRR processing.
  

Local fix

Problem summary

• There is a soft limit of 64MB on Direct Byte Buffers when using
  IBM JAVA ( we use it for cassandra ). Due to this soft limit,
  whenever client fetches large amount of data, cassandra errors
  out and rejects client connection with OODM ( Out of direct
  memory ).
  We encounter this error in all large fetches, for listing as
  well as aggregate queries.
  

Problem conclusion

• Fixes targeted for release v2018.4.1.3
  

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

  API CONNECT ENT

• Fixed component ID

  5725Z2201

Applicable component levels