apache/hadoop · warning · InterruptedIOException

Interrupted while waiting for IO on channel {}. Total timeou

Error message

Interrupted while waiting for IO on channel {}. Total timeout mills is {}, {} millis timeout left.

What it means

SelectorPool.select(long timeout) re-arms when select() spuriously returns 0 early, decrementing timeoutLeft each round, and after each round checks Thread.currentThread().isInterrupted(). An interrupt while parked is converted to InterruptedIOException carrying the total timeout and remaining budget — turning thread interruption into an IO-level cancellation signal for callers of doIO/connect/waitForIO.

Source

Thrown at hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/net/SocketIOWithTimeout.java:350

          long start = (timeout == 0) ? 0 : Time.now();

          key = channel.register(info.selector, ops);
          ret = info.selector.select(timeoutLeft);
          
          if (ret != 0) {
            return ret;
          }
          
          /* Sometimes select() returns 0 much before timeout for 
           * unknown reasons. So select again if required.
           */
          if (timeout > 0) {
            timeoutLeft -= Time.now() - start;
            timeoutLeft = Math.max(0, timeoutLeft);
          }
          
          if (Thread.currentThread().isInterrupted()) {
            throw new InterruptedIOException("Interrupted while waiting for "
                + "IO on channel " + channel + ". Total timeout mills is "
                + timeout + ", " + timeoutLeft + " millis timeout left.");
          }

          if (timeoutLeft == 0) {
            return 0;
          }
        }
      } finally {
        if (key != null) {
          key.cancel();
        }
        
        //clear the canceled key.
        try {
          info.selector.selectNow();
        } catch (IOException e) {
          LOG.info("Unexpected Exception while clearing selector : ", e);

View on GitHub (pinned to 2add963021)

Solutions

Treat InterruptedIOException as cancellation: abort the operation and release the channel/stream; do not blindly retry
Interrupt IO threads only as a last resort; prefer closing the channel, which unblocks the IO deterministically
If you catch it and swallow it, restore the flag with Thread.currentThread().interrupt() so upstream logic still sees the interrupt

Example fix

// before
executor.shutdownNow(); // interrupts IO threads mid-read -> opaque failure

// after
executor.shutdown();
if (!executor.awaitTermination(30, TimeUnit.SECONDS)) {
  channel.close(); // unblocks IO deterministically
  executor.shutdownNow();
}

Defensive patterns

Strategy: try-catch

Validate before calling

// no pre-check can prevent a concurrent interrupt; make cancellation cooperative
if (Thread.currentThread().isInterrupted()) throw new InterruptedIOException();

Try / catch

catch (InterruptedIOException e) {
  Thread.currentThread().interrupt(); // preserve the interrupt status
  abortTransferAndRelease();         // treat as cancellation, not a retryable error
}

Prevention

Cancel IO by closing the channel; it unblocks reads deterministically
Never swallow InterruptedIOException without restoring the interrupt flag
Avoid shutdownNow() on pools that own in-flight socket IO

When it happens

Trigger: Another thread interrupts the caller while it is blocked in a timed SocketIOWithTimeout read/write/connect: RPC call cancellation, ExecutorService.shutdownNow(), test teardown interrupting IO threads, or application shutdown hooks.

Common situations: Hadoop IPC clients cancelling calls; thread pools shutting down during block transfers; integration tests that interrupt worker threads instead of closing channels; frameworks that use interrupt as a general control signal.

Understand the failure class

Timeouts: ETIMEDOUT, deadlines, and hung requests — what actually expires when a request times out.

Related errors

AI-assisted analysis of apache/hadoop@2add963021 (2026-08-22). Data as JSON: /api/errors/acc4ab01247ec90d. Report an issue: GitHub.