Bazel Remote Cache Errors in CI - Fix Connection Failures

By Daniel Zoghalchali·Latchkey

Bazel could not talk to the remote cache or execution backend. A timeout, TLS/auth problem, or unreachable endpoint stops cache reads/writes. Transient connectivity issues clear on retry; with the right flags Bazel degrades gracefully instead of failing the build.

What this error means

The build prints remote-cache warnings or fails with "Failed to query remote execution capabilities" or repeated upload/download errors against the cache endpoint. Often intermittent and tied to the network or the cache service’s availability.

Bazel output

WARNING: Error reading from the remote cache:
  Failed to query remote execution capabilities:
  UNAVAILABLE: io exception
ERROR: Failed to fetch blobs because they do not exist remotely.

Common causes

Cache endpoint unreachable or slow

A network blip, DNS failure, or an overloaded cache backend makes the gRPC/HTTP cache endpoint time out, so reads and writes fail.

Auth or TLS misconfiguration

A missing/expired credential or a TLS trust problem rejects the connection to the remote cache or RBE service.

How to fix it

Let Bazel tolerate cache failures

Treat the remote cache as best-effort so a cache outage degrades to a local build instead of failing.

.bazelrc

build --remote_cache=grpcs://cache.example:443
build --remote_local_fallback
build --remote_upload_local_results=true
# do not hard-fail on cache errors
build --noremote_accept_cached --remote_retries=3

Fix auth and endpoint config

Verify the cache URL, scheme (grpc/grpcs/http/https), and port.
Refresh credentials/headers (--remote_header) and confirm TLS trust.
Raise --remote_timeout for slow links and add --remote_retries.

How to prevent it

Set --remote_local_fallback so cache outages do not fail builds.
Add --remote_retries and a sane --remote_timeout for flaky links.
Monitor cache backend availability and rotate credentials before expiry.

Frequently asked questions

What causes "Bazel remote cache error"?