"... S3ObjectInputStream, aborting HTTP connection" error

GeoServer GeoServer User
,
1

Hi

I’m running GeoServer v2.25.3 (oscarfonts Docker image) with S3 blobstore and just recently my production GeoServer logs are full of these errors, multiple times per second:

15 Jan 15:52:21 WARN [internal.S3AbortableInputStream] - Not all bytes were read from the S3ObjectInputStream, aborting HTTP connection. This is likely an error and may result in sub-optimal behavior. Request only the bytes you need via a ranged GET or drain the input stream after use.

Has anyone else seen this problem and know what causes it? The GeoServer appears to be working just fine, returning tiles as expected. No errors in AWS console.

Thanks

Peter

  • GeoServer Version2.25.3
  • Git Revision00835315c4cc1743ceee114f20d0e4e86cc181cf
  • Build Date19-Jul-2024 12:30
  • GeoTools Version31.3 (rev 5c48493594b3ba5197f94cfe3cce7073dc37a508)
  • GeoWebCache Version1.25-SNAPSHOT (rev 70b38c615f940b311e4756224a66013d6d84bb81/70b38c615f940b311e4756224a66013d6d84bb81)
2

Hi Peter,

Checking the release notes for GeoWebCache 1.25.x:

Shows 1.25.1 release note update with:

Align FileBlobStore with parameter handling practice of S3/Azure

The version you are running has 1.25-SNAPSHOT which indicates that GeoServer release was not quite done properly. It should be a specific GeoWebCache version, so it is easier to check release notes and be sure exactly what you are running.

Checking commit history Commits · GeoWebCache/geowebcache · GitHub

Shows S3 fix applied August 20th, since GeoServer 2.25.3 is released July 18th you are missing these recent fixes.

Please retest with a supported 2.26.1 or maintenance 2.25.5 release and let us know how it goes :slight_smile: