GoogleContainerTools/jib · error · CacheCorruptedException

Schema 1 manifests corrupted

Error message

Schema 1 manifests corrupted

What it means

For schema 1 (Docker v2.1) cached manifests, verifyImageMetadata requires that no manifest list was stored and that no config blobs are present, since schema 1 manifests embed their config differently. If either is present, the cached schema-1 metadata is inconsistent and CacheCorruptedException('Schema 1 manifests corrupted') is thrown.

Source

Thrown at jib-core/src/main/java/com/google/cloud/tools/jib/cache/CacheStorageReader.java:62

  @VisibleForTesting
  static void verifyImageMetadata(ImageMetadataTemplate metadata, Path metadataCacheDirectory)
      throws CacheCorruptedException {
    List<ManifestAndConfigTemplate> manifestsAndConfigs = metadata.getManifestsAndConfigs();
    if (manifestsAndConfigs.isEmpty()) {
      throw new CacheCorruptedException(metadataCacheDirectory, "Manifest cache empty");
    }
    if (manifestsAndConfigs.stream().anyMatch(entry -> entry.getManifest() == null)) {
      throw new CacheCorruptedException(metadataCacheDirectory, "Manifest(s) missing");
    }
    if (metadata.getManifestList() == null && manifestsAndConfigs.size() != 1) {
      throw new CacheCorruptedException(metadataCacheDirectory, "Manifest list missing");
    }

    ManifestTemplate firstManifest = manifestsAndConfigs.get(0).getManifest();
    if (firstManifest instanceof V21ManifestTemplate) {
      if (metadata.getManifestList() != null
          || manifestsAndConfigs.stream().anyMatch(entry -> entry.getConfig() != null)) {
        throw new CacheCorruptedException(metadataCacheDirectory, "Schema 1 manifests corrupted");
      }
    } else if (firstManifest instanceof BuildableManifestTemplate) {
      if (manifestsAndConfigs.stream().anyMatch(entry -> entry.getConfig() == null)) {
        throw new CacheCorruptedException(metadataCacheDirectory, "Schema 2 manifests corrupted");
      }
      if (metadata.getManifestList() != null
          && manifestsAndConfigs.stream().anyMatch(entry -> entry.getManifestDigest() == null)) {
        throw new CacheCorruptedException(metadataCacheDirectory, "Schema 2 manifests corrupted");
      }
    } else {
      throw new CacheCorruptedException(
          metadataCacheDirectory, "Unknown manifest type: " + firstManifest);
    }
  }

  private final CacheStorageFiles cacheStorageFiles;

  CacheStorageReader(CacheStorageFiles cacheStorageFiles) {

View on GitHub (pinned to fb949e2676)

Solutions

1. Clear the Jib cache directory and rebuild online so a single consistent metadata set is written
2. Prefer a base image served as schema 2/OCI (most modern registries); legacy schema 1 is deprecated by registries
3. Ensure only one build process uses the cache directory at a time (or use separate cache dirs per build)
4. Upgrade Jib to a current version, which handles registry manifest schema negotiation consistently

Example fix

// before
from openjdk:8   // very old tag, schema 1 registry response, stale cache
// after
rm -rf ~/.cache/google-cloud-tools-jib
<from><image>eclipse-temurin:17-jre</image></from>   // schema 2/OCI base image

crane manifest $BASE_IMAGE | jq -e '.schemaVersion == 2'

When it happens

Trigger: retrieveMetadata finds firstManifest instanceof V21ManifestTemplate while getManifestList() != null or any entry has getConfig() != null — schema-1 entries mixed with schema-2-style metadata in the cache.

Common situations: Base image registry serves schema 1 for very old images (e.g. legacy Docker Hub images) and the cache got mixed writes; cache written by an older Jib version then read by a newer one; concurrent builds writing different manifest schemas to the same cache.

Related errors

• Manifest(s) missing
• Layer file did not include valid hash: <layerFile>
• Manifest cache empty
• Manifest list missing
• Dependency required by the JAR (as specified in `Class-Path`

AI-assisted analysis of GoogleContainerTools/jib@fb949e2676 (2026-09-06). Data as JSON: /api/errors/b52ec42fcfad7b8d. Report an issue: GitHub.