The grpc_health_probe
utility allows you to query health of gRPC services that
expose service their status through the gRPC Health Checking Protocol.
grpc_health_probe
is meant to be used for health checking gRPC applications in
Kubernetes, using the exec probes.
โ ๏ธ Kubernetes has now built-in gRPC health checking capability as generally available. As a result, you might no longer need to use this tool and can use the native Kubernetes feature instead.This tool can still be useful if you are on older versions of Kubernetes, or using advanced configuration (such as custom metadata, TLS or finer timeout tuning), or not using Kubernetes at all.
This command-line utility makes a RPC to /grpc.health.v1.Health/Check
. If it
responds with a SERVING
status, the grpc_health_probe
will exit with
success, otherwise it will exit with a non-zero exit code (documented below).
EXAMPLES
$ grpc_health_probe -addr=localhost:5000
healthy: SERVING
$ grpc_health_probe -addr=localhost:9999 -connect-timeout 250ms -rpc-timeout 100ms
failed to connect service at "localhost:9999": context deadline exceeded
exit status 2
It is recommended to use a version-stamped binary distribution:
- Choose a binary from the Releases page.
Installing from source (not recommended):
- Make sure you have
git
andgo
installed. - Run:
go install github.com/grpc-ecosystem/grpc-health-probe@latest
- This will compile the binary into your
$GOPATH/bin
(or$HOME/go/bin
).
To make use of the grpc_health_probe
, your application must implement the
gRPC Health Checking Protocol v1. This means you must to register the
Health
service and implement the rpc Check
that returns a SERVING
status.
Since the Health Checking protocol is part of the gRPC core, it has packages/libraries available for the languages supported by gRPC:
[health.proto] [Go] [Java] [Python] [C#/NuGet] [Ruby] ...
Most of the languages listed above provide helper functions that hides
implementation details. This eliminates the need for you to implement the
Check
rpc yourself.
Kubernetes now supports gRPC health checking. If your cluster is running a version that supports gRPC health checking, you can define a gRPC liveness probe in your Pod specification. For more information on how to define a gRPC liveness probe in Kubernetes, see the Kubernetes documentation.
However, if your Kubernetes version does not support gRPC health checking or if you want to use some advanced features that Kubernetes does not support, you can use grpc_health_probe
to health-check your gRPC server. As a solution,
grpc_health_probe
can be used for Kubernetes to health-check gRPC
servers running in the Pod.
You are recommended to use Kubernetes exec
probes and define
liveness and/or readiness checks for your gRPC server pods.
You can bundle the statically compiled grpc_health_probe
in your container
image. Choose a binary release and download it in your Dockerfile:
RUN GRPC_HEALTH_PROBE_VERSION=v0.4.13 && \
wget -qO/bin/grpc_health_probe https://github.com/grpc-ecosystem/grpc-health-probe/releases/download/${GRPC_HEALTH_PROBE_VERSION}/grpc_health_probe-linux-amd64 && \
chmod +x /bin/grpc_health_probe
In your Kubernetes Pod specification manifest, specify a livenessProbe
and/or
readinessProbe
for the container:
spec:
containers:
- name: server
image: "[YOUR-DOCKER-IMAGE]"
ports:
- containerPort: 5000
readinessProbe:
exec:
command: ["/bin/grpc_health_probe", "-addr=:5000"]
initialDelaySeconds: 5
livenessProbe:
exec:
command: ["/bin/grpc_health_probe", "-addr=:5000"]
initialDelaySeconds: 10
This approach provide proper readiness/liveness checking to your applications that implement the gRPC Health Checking Protocol.
If a gRPC server is serving traffic over TLS, or uses TLS client authentication
to authorize clients, you can still use grpc_health_probe
to check health
with command-line options:
Option | Description |
---|---|
-tls |
use TLS (default: false) |
-tls-ca-cert |
path to file containing CA certificates (to override system root CAs) |
-tls-client-cert |
client certificate for authenticating to the server |
-tls-client-key |
private key for for authenticating to the server |
-tls-no-verify |
use TLS, but do not verify the certificate presented by the server (INSECURE) (default: false) |
-tls-server-name |
override the hostname used to verify the server certificate |
If your gRPC server requires authentication, you can use the following command line options and set the SPIFFE_ENDPOINT_SOCKET environment variable.
Option | Description |
---|---|
-spiffe |
use SPIFFE Workload API to retrieve TLS credentials (default: false) |
Option | Description |
---|---|
-v |
verbose logs (default: false) |
-connect-timeout |
timeout for establishing connection |
-rpc-timeout |
timeout for health check rpc |
-rpc-header |
sends metadata in the RPC request context (default: empty map) |
-user-agent |
user-agent header value of health check requests (default: grpc_health_probe) |
-service |
service name to check (default: "") - empty string is convention for server health |
-gzip |
use GZIPCompressor for requests and GZIPDecompressor for response (default: false) |
-version |
print the probe version and exit |
Example:
-
Start the
route_guide
example server with TLS by running:go run server/server.go -tls
-
Run
grpc_client_probe
with the CA certificate (in thetestdata/
directory) and hostname override the cert is signed for:$ grpc_health_probe -addr 127.0.0.1:10000 \ -tls \ -tls-ca-cert /path/to/testdata/ca.pem \ -tls-server-name=example.com \ -rpc-header=foo:bar \ -rpc-header=foo2:bar2 status: SERVING
It is not recommended to rely on specific exit statuses. Any failure will be a non-zero exit code.
Exit Code | Description |
---|---|
0 | success: rpc response is SERVING . |
1 | failure: invalid command-line arguments |
2 | failure: connection failed or timed out |
3 | failure: rpc failed or timed out |
4 | failure: rpc successful, but the response is not SERVING |
20 | failure: could not retrieve TLS credentials using the SPIFFE Workload API |
This is not an official Google project.