Apache Phoenix Known Issues

Potential deadlock on region opening with secondary index recovery

Distributed deadlock happens in clusters with a moderate number of regions for the data tables, secondary index tables.

Products affected: Phoenix

Releases affected:Phoenix 4.14.1 parcel

User affected: Users of Phoenix with secondary indexes

Severity (Low/Medium/High): Medium

Impact: Deadlock on region opening

Immediate action required/workaround:

Make the following configuration changes in HBase:

1) Set hbase.master.startup.retainassign to false

2) Increase the value of hbase.regionserver.executor.openregion.threads and restart cluster should bring up cluster.

If you continue to encounter issues, tune the following assignment manager parameters to match the count of regions for faster assignments:

  • hbase.assignment.threads.max
  • hbase.master.namespace.init.timeout
  • hbase.master.wait.on.regionservers.mintostart
  • hbase.bulk.assignment.threshold.regions
  • hbase.bulk.assignment.threshold.servers