Skip to main content

Troubleshooting

This article provides troubleshooting information for various tools and features in Pomerium.

Pomerium Core

HTTP Strict Transport Security (HSTS)

By default, Pomerium sends the Strict-Transport-Security response header to the browser, which pins the certificate to our browser for one year. This is common best practice to help prevent man-in-the-middle attacks but can create issues while a new Pomerium configuration is in development.

When you visit an endpoint while Pomerium is using an untrusted certificate (like the self-signed bootstrap certificate or a Let's Encrypt staging certificate), that certificate may be pinned in your browser. Once Pomerium is switched to a trusted production certificate, the untrusted cert must reset in the browser.

While developing your Pomerium environment, consider adjusting the SET_RESPONSE_HEADERS key to remove Strict-Transport-Security or reduce the max-age value until your production certificates are in place.

See this article for more information on clearing HSTS for specific endpoints across common browsers.

JWT Authentication

When securing the Pomerium Authenticate service with a certificate signed by Let's Encrypt, your upstream applications may reject the certificate when attempting to access the JWT signing key. Here's an example log line from Grafana:

logger=context error=Get "https://authenticate.localhost.pomerium.io/.well-known/pomerium/jwks.json": x509: certificate signed by unknown authority

This is often due to the recent expiration of the DST Root CA X3 certificate. Many default keystores used by docker images and less-frequently updated distributions still carry this expired certificate. Even though Let's Encrypt certs are cross-signed with the ISRG Root X1 CA certificate, some applications will still reject them.

To clarify; this does not mean that the upstream service is rejecting the JWT signing key. Rather, it doesn't trust the Let's Encrypt certificate used by the Authorize service for TLS, and so it will not read the JWKS file.

For upstream applications that can use a local signing key file, you can circumvent this issue using curl or wget to download the signing key locally (relative to the upstream service). Using Grafana again as an example:

  1. Download the jwks.json file from the authenticate server:

    curl https://authenticate.localhost.pomerium.io/.well-known/pomerium/jwks.json > /etc/grafana/jwks.json
  2. Edit the upstream service configuration to use the local key to verify tokens:

    [auth.jwt]
    enabled = true
    header_name = X-Pomerium-Jwt-Assertion
    email_claim = email
    jwk_set_file = /etc/grafana/jwks.json
    cache_ttl = 60m

Kubernetes Ingress Controller

View Event History

Pomerium Ingress Controller will add events to the Ingress objects as it processes them.

kubectl describe ingress/my-ingress
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Updated 18s pomerium-ingress updated pomerium configuration

If an error occurs, it may be reflected in the events:

Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Updated 5m53s pomerium-ingress updated pomerium configuration
Warning UpdateError 3s pomerium-ingress upsert routes: parsing ingress: annotations: applying policy annotations: parsing policy: invalid rules in policy: unsupported conditional "maybe", only and, or, not, nor and action are allowed

Shared Secret Mismatch

Pomerium's independent services communicate securely using a shared secret. When services or the databroker have mismatched secrets, Pomerium will fail.

Pomerium Core will log a shared secret mismatch with:

{
"level": "error",
"syncer_id": "authorize",
"syncer_type": "",
"error": "rpc error: code = Unauthenticated desc = invalid JWT: go-jose/go-jose: error in cryptographic primitive",
"time": "2022-03-22T07:26:14-04:00",
"message": "sync"
}

And Pomerium Enterprise will log the error with:

{
"level": "error",
"ts": "2022-03-22T07:21:02-04:00",
"caller": "dashboard/server.go:187",
"msg": "syncer",
"error": "failed to sync all devices: rpc error: code = Unauthenticated desc = invalid JWT: go-jose/go-jose: error in cryptographic primitive",
"stacktrace": "github.com/pomerium/pomerium-console/svc/dashboard.(*Server).Run.func2\n\t/PATH/TO/POMERIUM/CONSOLE/SERVICE/svc/dashboard/server.go:187\ngolang.org/x/sync/errgroup.(*Group).Go.func1\n\t/Users/tgroth/workspace/go/pkg/mod/golang.org/x/sync@v0.0.0-20210220032951-036812b2e83c/errgroup/errgroup.go:57"
}
{
"level": "info",
"ts": "2022-03-22T07:21:02-04:00",
"caller": "dashboard/server.go:202",
"msg": "stopping dashboard servers"
}

Update the shared secret across all Pomerium services to match the one set for the Databroker.

Redis Secret Mismatch

caution

The use of Redis as the databroker storage type has been deprecated in favor of Postgres. See Previous Versions of this page for information regarding Redis.

RPC Errors

certificate signed by unknown authority

When authenticating and authorizing a user for the first time, you may see the following in your Pomerium logs:

ERR http-error error="401 Unauthorized: ..... rpc error: code = DeadlineExceeded desc = latest connection error: connection error: desc = "transport: authentication handshake failed: x509: certificate signed by unknown authority...."

Why

This error means that the proxy is rejecting the Authorize service's supplied certificate (used to establish a secure connection) because it doesn't know or trust the certificate authority that signed it.

Solution

Ensure that the Proxy service knows about and trusts the certificate authority that signed the Authorize service's certificate.

  • Add the certificate authority directly into Pomerium using the certificate authority config setting.

  • Add the certificate authority to the system's underlying trust store.

  • Replace your system / docker image certificate bundle.

    For Docker:

    COPY --from=builder /etc/ssl/certs/your-cert-bundle.crt /etc/ssl/certs/ca-certificates.crt
  • Finally, ensure that you aren't being man-in-the-middle'd or that some eager router isn't injecting its own certificate along the way. Use openssl to verify that your Proxy service is getting the certificate you think its getting.

    openssl s_client -servername pomerium.io -connect pomerium.io:443 </dev/null \
    | sed -ne '/-BEGIN CERTIFICATE-/,/-END CERTIFICATE-/p'

rpc error: code = DeadlineExceeded

When authenticating and authorizing a user for the first time, you may get the following in your Pomerium logs.

{"level":"error",..."error":"rpc error: code = DeadlineExceeded desc = context deadline exceeded","http-message":"rpc error: code = DeadlineExceeded desc = context deadline exceeded","http-code":500,"message":"http-error"}

Why

The Proxy service is not able to create a connection with the authorization service to validate a user.

Solution

Usually, this is the result of either a routing issue or a configuration error. Make sure that you are using the internally routable URL for the Authorize service. Many cloud loud balancers do not yet support gRPC transposing the ingress. So while your authenticate service url will probably look like https://authenticate.corp.example.com, your authorizer service url will likely be more like https://pomerium-authorize-service.default.svc.cluster.local or https://localhost:5443.

Pomerium Enterprise

Generate Recovery Token

In the event that you lose access to the console via delegated access (the policy defined in Pomerium), there exists a fallback procedure to regain access to the console via a generated recovery token.

Pomerium Enterprise Recovery Sign In

To generate a token, run the pomerium-console generate-recovery token command with the following flags:

FlagDescription
--database-encryption-keybase64-encoded encryption key for encrypting sensitive data in the database.
--database-urlThe database to connect to (default "postgresql://pomerium:pomerium@localhost:5432/dashboard?sslmode=disable").
--namespaceThe namespace to use (default "9d8dbd2c-8cce-4e66-9c1f-c490b4a07243" for Global).
--outWhere to save the JWT. If not specified, it will be printed to stdout.
--ttlThe amount of time before the recovery token expires. Requires a unit (example: 30s, 5m).
tip

You can run the pomerium-console binary from any device with access to the database.

Envoy error messages

Because Pomerium relies on Envoy to manage HTTP connections, you will notice Envoy connection errors and messages at some point in your logs as you configure Pomerium.

The Envoy Response Code Details provides an exhaustive list of Envoy-related message details.

We've repurposed a truncated version of the Response Code Details list in the table below for your convenience:

NameDescription
absolute_path_rejectedThe request was rejected due to using an absolute path on a route not supporting them.
admin_filter_responseThe response was generated by the admin filter.
cluster_not_foundThe request was rejected by the router filter because there was no cluster found for the selected route.
downstream_local_disconnectThe client connection was locally closed for the provided reason.
downstream_remote_disconnectThe client disconnected unexpectedly.
duration_timeoutThe max connection duration was exceeded.
direct_responseA direct response was generated by the router filter.
filter_added_invalid_request_dataA filter added request data at the wrong stage in the filter chain.
filter_added_invalid_response_dataA filter added response data at the wrong stage in the filter chain.
filter_chain_not_foundThe request was rejected due to no matching filter chain.
filter_removed_required_request_headersThe request was rejected in the filter manager because a configured filter removed required request headers.
filter_removed_required_response_headersThe response was rejected in the filter manager because a configured filter removed required response headers or these values were invalid (e.g. overflown status).
internal_redirectThe original stream was replaced with an internal redirect.
low_versionThe HTTP/1.0 or HTTP/0.9 request was rejected due to HTTP/1.0 support not being configured.
maintenance_modeThe request was rejected by the router filter because the cluster was in maintenance mode.
max_duration_timeoutThe per-stream max duration timeout was exceeded.
missing_host_headerThe request was rejected due to a missing Host: or :authority field.
missing_path_rejectedThe request was rejected due to a missing Path or :path header field.
no_healthy_upstreamThe request was rejected by the router filter because there was no healthy upstream found.
overloadThe request was rejected due to the Overload Manager reaching configured resource limits.
rejecting_because_detection_failedThe request was rejected because the original IP couldn’t be detected.
path_normalization_failedThe request was rejected because path normalization was configured on and failed, probably due to an invalid path.
request_headers_failed_strict_checkThe request was rejected due to x-envoy-* headers failing strict header validation.
request_overall_timeoutThe per-stream total request timeout was exceeded.
request_payload_exceeded_retry_buffer_limitEnvoy is doing streaming proxying but too much data arrived while waiting to attempt a retry.
request_payload_too_largeEnvoy is doing non-streaming proxying and the request payload exceeded configured limits.
response_payload_too_largeEnvoy is doing non-streaming proxying and the response payload exceeded configured limits.
route_configuration_not_foundThe request was rejected because there was no route configuration found.
route_not_foundThe request was rejected because there was no route found.
stream_idle_timeoutThe per-stream keepalive timeout was exceeded.
upgrade_failedThe request was rejected because it attempted an unsupported upgrade.
upstream_max_stream_duration_reachedThe request was destroyed because of it exceeded the configured max stream duration.
upstream_per_try_timeoutThe final upstream try timed out.
upstream_reset_after_response_started{details}The upstream connection was reset after a response was started. This may include further details about the cause of the disconnect.
upstream_reset_before_response_started{details}The upstream connection was reset before a response was started This may include further details about the cause of the disconnect.
upstream_response_timeoutThe upstream response timed out.
via_upstreamThe response code was set by the upstream.

Miscellaneous

Invalid Certificates from Command Line Tools

When using Let's Encrypt certificates, you must use the fullchain.pem file, not cert.pem in order to include intermediate certs. Browsers like Chrome will store intermediate certs for LE but other tools (like curl) don't, which is why your route might look fine in a web browser, but not when curl'd or used for TCP tunneling.