Phoenix global secondary index stuck in FAILED state

When creating a global secondary index in Apache Phoenix, the system registers a PENDING entry in the SYSTEM.CATALOG and initiates a background task to populate data across HBase regions. If a region server hosting a specific partition becomes unavailable during this process, the build task may abort, leaving the index in a FAILED state.

This failure state is not surfaced directly through the Phoenix client-side API, and the specific root cause is contained only within the region server or query server logs. Because an index in a FAILED state can block subsequent DDL operations on the parent table, identifying the exact region-level failure is necessary for recovery.

Diagnostic Requirements

• What are the specific log patterns or error codes in Phoenix 4.x and 5.x that indicate an index build abort?
• Is there an automated retry mechanism for failed index builds, or is a manual drop and recreation required every time?