What causes "Asked to load path[%s], but it doesn't exist." in apache/druid?

Check the path in the message on the historical node; restore the file from backup or another replica.

apache/druid · error · SegmentLoadingException

Asked to load path[%s], but it doesn't exist.

Error message

Asked to load path[%s], but it doesn't exist.

What it means

LocalDataSegmentPuller.getFile throws SegmentLoadingException when the path in the segment's local loadSpec does not exist on the filesystem. Druid was asked to load a segment whose recorded deep-storage file is missing, so it cannot serve data from this copy.

Source

Thrown at server/src/main/java/org/apache/druid/segment/loading/LocalDataSegmentPuller.java:241

    return new Predicate<>()
    {
      @Override
      public boolean apply(Throwable input)
      {
        return !(input instanceof InterruptedException)
               && !(input instanceof CancellationException)
               && (input instanceof Exception);
      }
    };
  }

  private File getFile(DataSegment segment) throws SegmentLoadingException
  {
    final Map<String, Object> loadSpec = segment.getLoadSpec();
    final File path = new File(MapUtils.getString(loadSpec, "path"));

    if (!path.exists()) {
      throw new SegmentLoadingException("Asked to load path[%s], but it doesn't exist.", path);
    }

    return path;
  }
}

View on GitHub (pinned to 9b90983fd2)

Solutions

Check the path in the message on the historical node; restore the file from backup or another replica.
Verify druid.storage.baseDir/baseKeyPrefix config matches where segments were actually pushed.
Kill the missing segment via the coordinator kill API (or disable it) so metadata stops referencing it.
If segments were moved, re-push them or update the loadSpec paths in metadata store to the new location.

Example fix

// before: base dir changed in config but old segments still on old path
# druid.storage.baseDir=/mnt/new/deepstorage  (segments live under /mnt/deepstorage)
// after: point config back or move data
# druid.storage.baseDir=/mnt/deepstorage
mv /mnt/deepstorage/druid/segments /mnt/new/deepstorage/druid/segments

Defensive patterns

Strategy: validation

Validate before calling

String path = (String) segment.getLoadSpec().get("path");
if (path == null || !new File(path).exists()) {
  throw new IllegalStateException("segment data missing on disk: " + path);
}

Type guard

boolean localSegmentExists(DataSegment segment) {
  Object p = segment.getLoadSpec().get("path");
  return p instanceof String && new File((String) p).exists();
}

Try / catch

try {
  File f = puller.getSegmentFiles(segment, outDir);
} catch (SegmentLoadingException e) {
  log.warn(e, "local segment file missing for %s; triggering re-replication", segment.getId());
}

Prevention

Don't delete or prune files in Druid deep storage with ad-hoc cron jobs; use the coordinator kill API.
Keep druid.storage.baseDir/baseKeyPrefix consistent across cluster config changes.
Use a shared, persistent filesystem for local deep storage across historical restarts.
Maintain >=2 replicas so missing copies can be re-replicated automatically.

When it happens

Trigger: getFile(DataSegment) resolves loadSpec 'path' via new File(...) and path.exists() returns false — file deleted, host migrated, or path recorded on a different machine.

Common situations: Deep-storage directory wiped or cleaned by a retention job; local storage shared-name mismatch after host move; restore/archive operations removed the file but metadata still references it; typo in druid.storage.baseDir between clusters.

Understand the failure class

Background: "File not found" and ENOENT errors: why libraries can't find a file that should exist — this error's family across 50 libraries.

Related errors

AI-assisted analysis of apache/druid@9b90983fd2 (2026-09-07). Data as JSON: /api/errors/70c9a7a22c8e4285. Report an issue: GitHub.