`actions/cache@v4` save fails: “Unable to extract blobName and containerName …” with github-actions-cache-server (sqlite/filesystem driver)

[paulharkink]

Summary

actions/cache/save@v4 consistently fails with:

Warning: Failed to save: Unable to extract blobName and containerName with provided information.

---

Environment

• Kubernetes (self‑hosted GitHub Actions runners via ARC runner scale set)
• gha-runner-scale-set-controller chart: 0.13.1
• gha-runner-scale-set chart: 0.13.1
• Runner image: ghcr.io/falcondev-oss/actions-runner:2.331.0
• Cache server chart: ghcr.io/falcondev-oss/charts/github-actions-cache-server:9.2.0
• Cache server storage: filesystem + sqlite database on a PVC

---

Runner configuration

The runner pods are configured with:

env:
  - name: CUSTOM_ACTIONS_RESULTS_URL
    value: "http://gha-cache-server-github-actions-cache-server.arc-systems.svc.cluster.local/"

Per the chart documentation, the patched runner is expected to derive ACTIONS_RESULTS_URL internally from CUSTOM_ACTIONS_RESULTS_URL. I am intentionally not setting ACTIONS_RESULTS_URL or ACTIONS_CACHE_URL manually.

---

Reproduction workflow

Minimal workflow demonstrating the failure:

jobs:
  cache-test:
    runs-on: myCluster
    steps:
      - uses: actions/checkout@v4

      - name: Create file to cache
        run: |
          mkdir -p build/test
          date > build/test/hello.txt

      - uses: actions/cache/save@v4
        with:
          path: |
            build
          key: build-outputs-${{ github.run_id }}-${{ github.run_attempt }}

---

Logs (save step)

/usr/bin/tar --posix -cf cache.tgz --exclude cache.tgz -P -C ... --files-from manifest.txt -z
Warning: Failed to save: Unable to extract blobName and containerName with provided information.
Warning: Cache save failed.

---

Additional observations

• When using gradle/actions/setup-gradle@v4, I sometimes see:

  ReserveCacheError: Unable to reserve cache with key ..., another job may be creating this cache.

• However, inspecting the sqlite DB on the cache-server PVC shows no cache entries being recorded:

sqlite> select * from kysely_migration;
$0_init|2026-02-20T16:51:58.851Z
sqlite> select * from kysely_migration_lock;
migration_lock|0

No cache records appear to be written even though reserve attempts are reported in job logs.

---

Question

Is the above configuration expected to support actions/cache@v4 save operations?

Specifically:

• Is CUSTOM_ACTIONS_RESULTS_URL alone sufficient configuration for this chart and runner image?
• Are additional environment variables required for the cache v4 protocol?
• Does this error indicate a mismatch with the newer cache protocol implementation used by actions/cache@v4?

Any guidance on the correct runner environment configuration for cache save support would be appreciated.

[paulharkink]

I did some controlled testing to isolate what’s happening.

If I point CUSTOM_ACTIONS_RESULTS_URL to a non-existent host, the cache step just ends with a generic:

Cache save failed.

If I point it to the cache-server service, the error consistently becomes:

Failed to save: Unable to extract blobName and containerName with provided information.

So the runner is definitely reaching the cache server and getting a response — the error only appears when talking to it.

[LouisHaftmann]

I can't reproduce the issues you mentioned. Could you try running the cache server with DEBUG=true and sharing the logs?