Maintenance eviction pass failed; retired connections…

Error message

Maintenance eviction pass failed; retired connections recycle on return

What it means

ConnectionPool.evictQuietly runs the pool's maintenance eviction pass (evict()) in the background and deliberately swallows any exception, logging this warning. When the pass fails, retired/marking connections are not proactively evicted; they are instead recycled on the next return to the pool. The library continues running — this is degraded proactive maintenance, not a request failure.

Solutions

Inspect the logged cause (full stack trace accompanies this warning) and address the underlying connection failure.
Ensure the pool is not being closed while background maintenance runs.
Verify Redis availability; if the server is down, the warnings are expected until it returns.
As a safety net, confirm connections validate on borrow so stale ones are destroyed on use even if eviction fails.

Example fix

// before
pool = new ConnectionPool(poolConfig, factory); // maintenance thread logs repeated warnings when Redis is down
// after
if (factory instanceof ConnectionFactory && !redisReachable()) {
  // suspend maintenance or back off instead of letting evict() throw every cycle
  poolConfig.setTestWhileIdle(false);
  pool = new ConnectionPool(poolConfig, factory);
}

Defensive patterns

Strategy: retry

Validate before calling

if (!pool.isClosed()) {
  pool.getResource().ping(); // sanity check before relying on maintenance passes
}

Try / catch

try { pool.evict(); } catch (Exception e) { logger.warn("eviction pass failed, retrying later", e); }

Prevention

Don't close the pool while its maintenance thread is running
Fix root network instability — repeated warnings usually mean Redis is down or flapping
Rely on testOnBorrow as a backstop when idle eviction fails
Track the logged cause stack trace to the actual failing connection check

When it happens

Trigger: The internal evict() pass throws while scanning idle objects — typically because an underlying connection check (PING/socket read) throws an unexpected exception, or the pool is in an inconsistent state during shutdown.

Common situations: Network instability causing eviction-time idle checks to throw; pool being closed concurrently with the maintenance thread; Redis going down so every idle-object check raises an exception repeatedly.

Related errors

AI-assisted analysis of redis/jedis@6dac31d4c2 (2026-09-08). Data as JSON: /api/errors/79a2fadec685aa29. Report an issue: GitHub.

Appendix: source

Thrown at src/main/java/redis/clients/jedis/ConnectionPool.java:126

      };
    } else {
      returnHook = super::returnResource;
    }
  }

  /**
   * Handoff-hook reaction: evict retired idles. Runs on the maintenance scheduler thread or inline
   * on a notifying thread; must never propagate (a failed pass degrades to lazy recycling on
   * return).
   */
  private void evictQuietly() {
    if (isClosed()) {
      return;
    }
    try {
      evict();
    } catch (Exception e) {
      log.warn("Maintenance eviction pass failed; retired connections recycle on return", e);
    }
  }

  /** Exposes the pool's maintenance controller ({@code null} when off) for test clock injection. */
  @VisibleForTesting
  MaintenanceEventController getMaintenanceController() {
    return maintenanceController;
  }

  @Override
  public Connection getResource() {
    Connection conn = super.getResource();
    conn.setHandlingPool(this);
    return conn;
  }

  @Override
  public void close() {

View on GitHub (pinned to 6dac31d4c2)