IJ10680: RECURRENT ABORTED SCAVENGE

APAR status

• Closed as program error.

Error description

• Error Message: Reduced performance or even OOM due to excessive
  GC, due to frequent aborted Scavenge.
  .
  Stack Trace: N/A
  .
  Frequent series of events when percolated global GC caused by an
  aborted Scavenge, does not end up with successful allocation
  from Nursery, but from Tenure. Also, subsequent allocations
  would trigger Scavenge GCs, even though Nursery has plenty of
  free space.
  <gc-op id="1111" type="scavenge" timems="1875.393"
  contextid="1108" timestamp="2018-09-21T16:28:41.557">
    <scavenger-info tenureage="5" tenuremask="fde0" tiltratio="81"
  />
    <memory-copied type="nursery" objects="668" bytes="268578080"
  bytesdiscarded="805258064" />
    <memory-copied type="tenure" objects="1" bytes="268323976"
  bytesdiscarded="0" />
    <copy-failed type="nursery" objects="1" bytes="268323976" />
    <copy-failed type="tenure" objects="4" bytes="1073295904" />
    <ownableSynchronizers candidates="292" cleared="0" />
    <warning details="aborted collection due to insufficient free
  space" />
  </gc-op>
  ....
  <percolate-collect id="1115" from="nursery" to="global"
  reason="failed tenure threshold reached"
  timestamp="2018-09-21T16:28:41.557"/>
  ....
  <af-end id="1126" timestamp="2018-09-21T16:28:41.580"
  threadId="00007FFC981C1CE0" success="true" from="tenure-soa"/>
                  <<< normally should come from Nursery
  <exclusive-end id="1127" timestamp="2018-09-21T16:28:41.580"
  durationms="1899.036" />
  ....
  <af-start id="1129" threadId="000000000241DFE0"
  totalBytesRequested="24" timestamp="2018-09-21T16:28:41.581"
  intervalms="1899.179" type="nursery" />
        <mem type="allocate" free="4949081512" total="5501747200"
  percent="89" />
                                                  <<< GC
  triggered, but plenty of free memory
  

Local fix

• The problem could be somewhat reduced by increasing heap size,
  especially Tenure space.
  

Problem summary

• After a percolate global GC which follows an aborted Scavenge,
  further Nursery allocation would be incorrectly inhibited. The
  problem would often be implicitly cleared by next Scavenge
  itself, but in more extreme case, this would lead to a series of
  aborted and percolate global GCs.
  

Problem conclusion

• Percolate global GC is fixed to allow Nursery allocations.
  .
  This APAR will be fixed in the following Java Releases:
     8    SR5 FP25  (8.0.5.25)
  .
  Contact your IBM Product's Service Team for these Service
  Refreshes and Fix Packs.
  For those running stand-alone, information about the available
  Service Refreshes and Fix Packs can be found at:
             https://www.ibm.com/developerworks/java/jdk/
  

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

  J9 COMMON CODE

• Fixed component ID

  620700127

Applicable component levels

• R270 PSY

     UP