Consul includes a built-in tool for troubleshooting communication between services in a service mesh. Learn how to use the `consul troubleshoot` command to validate communication between upstream and downstream Envoy proxies on VM and Kubernetes deployments.
---
# Service-to-service troubleshooting overview
This topic provides an overview of Consul’s built-in service-to-service troubleshooting capabilities. When communication between an upstream service and a downstream service in a service mesh fails, you can run the `consul troubleshoot` command to initiate a series of automated validation tests.
For more information, refer to the [`consul troubleshoot` CLI documentation](/consul/commands/troubleshoot) or the [`consul-k8s troubleshoot` CLI reference](/consul/docs/k8s/k8s-cli#troubleshoot).
When communication between upstream and downstream services in a service mesh fails, you can diagnose the cause manually with one or more of Consul’s built-in features, including [health check queries](/consul/docs/services/usage/checks), [the UI topology view](/consul/docs/connect/observability/ui-visualization), and [agent telemetry metrics](/consul/docs/agent/telemetry#metrics-reference).
The `consul troubleshoot` command performs several checks in sequence that enable you to discover issues that impede service-to-service communication. The process systematically queries the [Envoy administration interface API](https://www.envoyproxy.io/docs/envoy/latest/operations/admin) and the Consul API to determine the cause of the communication failure.
The troubleshooting command validates service-to-service communication by checking for the following common issues:
- Upstream service does not exist
- One or both hosts are unhealthy
- A filter affects the upstream service
- The CA has expired mTLS certificates
- The services have expired mTLS certificates
Consul outputs the results of these validation checks to the terminal along with suggested actions to resolve the service communication failure. When it detects rejected configurations or connection failures, Consul also outputs Envoy metrics for services.
### Envoy proxies in a service mesh
Consul validates communication in a service mesh by checking the Envoy proxies that are deployed as sidecars for the upstream and downstream services. As a result, troubleshooting requires that [Consul’s service mesh features are enabled](/consul/docs/connect/configuration).
For more information about using Envoy proxies with Consul, refer to [Envoy proxy configuration for service mesh](/consul/docs/connect/proxies/envoy).
## Requirements
- Consul v1.15 or later.
- For Kubernetes, the `consul-k8s` CLI must be installed.
### Technical constraints
When troubleshooting service-to-service communication issues, be aware of the following constraints:
- The troubleshooting tool does not check service intentions. For more information about intentions, including precedence and match order, refer to [service mesh intentions](/consul/docs/connect/intentions).
- The troubleshooting tool validates one direct connection between a downstream service and an upstream service. You must run the `consul troubleshoot` command with the Envoy ID for an individual upstream service. It does support validating multiple connections simultaneously.
- The troubleshooting tool only validates Envoy configurations for sidecar proxies. As a result, the troubleshooting tool does not validate Envoy configurations on upstream proxies such as mesh gateways and terminating gateways.
## Usage
Using the service-to-service troubleshooting tool is a two-step process:
1. Find the identifier for the upstream service.
1. Use the upstream’s identifier to validate communication.
In deployments without transparent proxies, the identifier is the _Envoy ID for the upstream service’s sidecar proxy_. If you use transparent proxies, the identifier is the _upstream service’s IP address_. For more information about using transparent proxies, refer to [Enable transparent proxy mode](/consul/docs/connect/transparent-proxy).
### Troubleshoot on VMs
To troubleshoot service-to-service communication issues in deployments that use VMs or bare-metal servers:
1. Run the `consul troubleshoot upstreams` command to retrieve the upstream information for the service that is experiencing communication failures. Depending on your network’s configuration, the upstream information is either an Envoy ID or an IP address.
If you cannot find the upstream address or cluster for a transparent proxy upstream:
- Check intentions: Tproxy upstreams are configured based on intentions. Make sure you have configured intentions to allow traffic to your upstream.
- To check that the right cluster is being dialed, run a DNS lookup for the upstream you are dialing. For example, run `dig backend.svc.consul` to return the IP address for the `backend` service. If the address you get from that is missing from the upstream IPs, it means that your proxy may be misconfigured.
```
1. Run the `consul troubleshoot proxy` command and specify the Envoy ID or IP address with the `-upstream-ip` flag to identify the proxy you want to perform the troubleshooting process on. The following example uses the upstream IP to validate communication with the upstream service `backend`:
```shell-session
$ consul troubleshoot proxy -upstream-ip 10.4.6.160
==> Validation
✓ Certificates are valid
✓ Envoy has 0 rejected configurations
✓ Envoy has detected 0 connection failure(s)
✓ Listener for upstream "backend" found
✓ Route for upstream "backend" found
✓ Cluster "backend.default.dc1.internal.e08fa6d6-e91e-dfe0-f6e1-ba097a828e31.consul" for upstream "backend" found
✓ Healthy endpoints for cluster "backend.default.dc1.internal.e08fa6d6-e91e-dfe0-f6e1-ba097a828e31.consul" for upstream "backend" found
✓ Cluster "backend2.default.dc1.internal.e08fa6d6-e91e-dfe0-f6e1-ba097a828e31.consul" for upstream "backend" found
! No healthy endpoints for cluster "backend2.default.dc1.internal.e08fa6d6-e91e-dfe0-f6e1-ba097a828e31.consul" for upstream "backend" found
-> Check that your upstream service is healthy and running
-> Check that your upstream service is registered with Consul
-> Check that the upstream proxy is healthy and running
-> If you are explicitly configuring upstreams, ensure the name of the upstream is correct
```
In the example output, troubleshooting upstream communication reveals that the `backend` service has two service instances running in datacenter `dc1`. One of the services is healthy, but Consul cannot detect healthy endpoints for the second service instance. This information appears in the following lines of the example:
```text hideClipboard
✓ Cluster "backend.default.dc1.internal.e08fa6d6-e91e-dfe0-f6e1-ba097a828e31.consul" for upstream "backend" found
✓ Healthy endpoints for cluster "backend.default.dc1.internal.e08fa6d6-e91e-dfe0-f6e1-ba097a828e31.consul" for upstream "backend" found
✓ Cluster "backend2.default.dc1.internal.e08fa6d6-e91e-dfe0-f6e1-ba097a828e31.consul" for upstream "backend" found
! No healthy endpoints for cluster "backend2.default.dc1.internal.e08fa6d6-e91e-dfe0-f6e1-ba097a828e31.consul" for upstream "backend" found
The output from the troubleshooting process identifies service instances according to their [Consul DNS address](/consul/docs/services/discovery/dns-static-lookups#standard-lookup). Use the DNS information for failing services to diagnose the specific issues affecting the service instance.
For more information, refer to the [`consul troubleshoot` CLI documentation](/consul/commands/troubleshoot).
### Troubleshoot on Kubernetes
To troubleshoot service-to-service communication issues in deployments that use Kubernetes, retrieve the upstream information for the pod that is experiencing communication failures and use the upstream information to identify the proxy you want to perform the troubleshooting process on.
1. Run the `consul-k8s troubleshoot upstreams` command and specify the pod ID with the `-pod` flag to retrieve upstream information. Depending on your network’s configuration, the upstream information is either an Envoy ID or an IP address. The following example displays all transparent proxy upstreams in Consul service mesh from the given pod.
If you cannot find the upstream address or cluster for a transparent proxy upstream:
- Check intentions: Tproxy upstreams are configured based on intentions. Make sure you have configured intentions to allow traffic to your upstream.
- To check that the right cluster is being dialed, run a DNS lookup for the upstream you are dialing. For example, run `dig backend.svc.consul` to return the IP address for the `backend` service. If the address you get from that is missing from the upstream IPs, it means that your proxy may be misconfigured.
```
1. Run the `consul-k8s troubleshoot proxy` command and specify the pod ID and upstream IP address to identify the proxy you want to troubleshoot. The following example uses the upstream IP to validate communication with the upstream service `backend`:
✓ cluster "backend.default.dc1.internal..consul" for upstream "backend" found
✓ healthy endpoints for cluster "backend.default.dc1.internal.e08fa6d6-e91e-dfe0-f6e1-ba097a828e31.consul" for upstream "backend" found
✓ cluster "backend2.default.dc1.internal..consul" for upstream "backend" found
! no healthy endpoints for cluster "backend2.default.dc1.internal.e08fa6d6-e91e-dfe0-f6e1-ba097a828e31.consul" for upstream "backend" found
```
In the example output, troubleshooting upstream communication reveals that the `backend` service has two clusters in datacenter `dc1`. One of the clusters returns healthy endpoints, but Consul cannot detect healthy endpoints for the second cluster. This information appears in the following lines of the example:
```text hideClipboard
✓ cluster "backend.default.dc1.internal..consul" for upstream "backend" found
✓ healthy endpoints for cluster "backend.default.dc1.internal.e08fa6d6-e91e-dfe0-f6e1-ba097a828e31.consul" for upstream "backend" found
✓ cluster "backend2.default.dc1.internal..consul" for upstream "backend" found
! no healthy endpoints for cluster "backend2.default.dc1.internal.e08fa6d6-e91e-dfe0-f6e1-ba097a828e31.consul" for upstream "backend" found
```
The output from the troubleshooting process identifies service instances according to their [Consul DNS address](/consul/docs/k8s/dns). Use the DNS information for failing services to diagnose the specific issues affecting the service instance.
For more information, refer to the [`consul-k8s troubleshoot` CLI reference](/consul/docs/k8s/k8s-cli#troubleshoot).
