There could be several reasons to why a sidecar will not be injected into a pod. First, check your deployment or pod YAML file, and check that you have the following annotations in the right place:
annotations:
dapr.io/enabled: "true"
dapr.io/app-id: "nodeapp"
dapr.io/app-port: "3000"
apiVersion: apps/v1
kind: Deployment
metadata:
name: nodeapp
namespace: default
labels:
app: node
spec:
replicas: 1
selector:
matchLabels:
app: node
template:
metadata:
labels:
app: node
annotations:
dapr.io/enabled: "true"
dapr.io/app-id: "nodeapp"
dapr.io/app-port: "3000"
spec:
containers:
- name: node
image: dapriosamples/hello-k8s-node
ports:
- containerPort: 3000
imagePullPolicy: Always
If your pod spec template is annotated correctly and you still don’t see the sidecar injected, make sure Dapr was deployed to the cluster before your deployment or pod were deployed.
If this is the case, restarting the pods will fix the issue.
If you are deploying Dapr on a private GKE cluster, sidecar injection does not work without extra steps. See Setup a Google Kubernetes Engine cluster.
In order to further diagnose any issue, check the logs of the Dapr sidecar injector:
kubectl logs -l app=dapr-sidecar-injector -n dapr-system
Note: If you installed Dapr to a different namespace, replace dapr-system above with the desired namespace
If the Dapr sidecar (daprd
) is taking too long to initialize, this might be surfaced as a failing health check by Kubernetes.
If your pod is in a failed state you should check this:
kubectl describe pod <name-of-pod>
You might see a table like the following at the end of the command output:
Normal Created 7m41s (x2 over 8m2s) kubelet, aks-agentpool-12499885-vmss000000 Created container daprd
Normal Started 7m41s (x2 over 8m2s) kubelet, aks-agentpool-12499885-vmss000000 Started container daprd
Warning Unhealthy 7m28s (x5 over 7m58s) kubelet, aks-agentpool-12499885-vmss000000 Readiness probe failed: Get http://10.244.1.10:3500/v1.0/healthz: dial tcp 10.244.1.10:3500: connect: connection refused
Warning Unhealthy 7m25s (x6 over 7m55s) kubelet, aks-agentpool-12499885-vmss000000 Liveness probe failed: Get http://10.244.1.10:3500/v1.0/healthz: dial tcp 10.244.1.10:3500: connect: connection refused
Normal Killing 7m25s (x2 over 7m43s) kubelet, aks-agentpool-12499885-vmss000000 Container daprd failed liveness probe, will be restarted
Warning BackOff 3m2s (x18 over 6m48s) kubelet, aks-agentpool-12499885-vmss000000 Back-off restarting failed container
The message Container daprd failed liveness probe, will be restarted
indicates at the Dapr sidecar has failed its health checks and will be restarted. The messages Readiness probe failed: Get http://10.244.1.10:3500/v1.0/healthz: dial tcp 10.244.1.10:3500: connect: connection refused
and Liveness probe failed: Get http://10.244.1.10:3500/v1.0/healthz: dial tcp 10.244.1.10:3500: connect: connection refused
show that the health check failed because no connection could be made to the sidecar.
The most common cause of this failure is that a component (such as a state store) is misconfigured and is causing initialization to take too long. When initialization takes a long time, it’s possible that the health check could terminate the sidecar before anything useful is logged by the sidecar.
To diagnose the root cause:
Remember to configure the liveness check delay and log level back to your desired values after solving the problem.
Have you installed an Dapr State store in your cluster?
To check, use kubectl get a list of components:
kubectl get components
If there isn’t a state store component, it means you need to set one up. Visit here for more details.
If everything’s set up correctly, make sure you got the credentials right. Search the Dapr runtime logs and look for any state store errors:
kubectl logs <name-of-pod> daprd
Have you installed an Dapr Message Bus in your cluster?
To check, use kubectl get a list of components:
kubectl get components
If there isn’t a pub/sub component, it means you need to set one up. Visit here for more details.
If everything is set up correctly, make sure you got the credentials right. Search the Dapr runtime logs and look for any pub/sub errors:
kubectl logs <name-of-pod> daprd
Check that there’s only one installation of the Dapr Operator in your cluster. Find out by running
kubectl get pods -l app=dapr-operator --all-namespaces
If two pods appear, delete the redundant Dapr installation.
This means there are some internal issue inside the Dapr runtime. To diagnose, view the logs of the sidecar:
kubectl logs <name-of-pod> daprd
This means you’re trying to call an Dapr API endpoint that either doesn’t exist or the URL is malformed. Look at the Dapr API reference here and make sure you’re calling the right endpoint.
Have you specified the port your app is listening on? In Kubernetes, make sure the dapr.io/app-port
annotation is specified:
annotations:
dapr.io/enabled: "true"
dapr.io/app-id: "nodeapp"
dapr.io/app-port: "3000"
If using Dapr Standalone and the Dapr CLI, make sure you pass the --app-port
flag to the dapr run
command.
The first thing to do is inspect the HTTP error code returned from the Dapr API, if any. If you still can’t find the issue, try enabling debug
log levels for the Dapr runtime. See here how to do so.
You might also want to look at error logs from your own process. If running on Kubernetes, find the pod containing your app, and execute the following:
kubectl logs <pod-name> <name-of-your-container>
If running in Standalone mode, you should see the stderr and stdout outputs from your app displayed in the main console session.
Each Dapr instance reports it’s host address to the placement service. The placement service then distributes a table of nodes and their addresses to all Dapr instances. If that host address is unreachable, you are likely to encounter socket timeout errors or other variants of failing request errors.
Unless the host name has been specified by setting an environment variable named DAPR_HOST_IP
to a reachable, pingable address, Dapr will loop over the network interfaces and select the first non-loopback address it finds.
As described above, in order to tell Dapr what the host name should be used, simply set an environment variable with the name of DAPR_HOST_IP
.
The following example shows how to set the Host IP env var to 127.0.0.1
:
Note: for versions <= 0.4.0 use HOST_IP
export DAPR_HOST_IP=127.0.0.1
This is usually due to one of the following issues
NAMESPACE
environment variable locally or deployed your components into a different namespace in Kubernetes. Check which namespace your app and the components are deployed to. Read scoping components to one or more applications for more information.--components-path
with the Dapr run
commands or not placed your components into the default components folder for your OS. Read define a component for more information.Some organizations will implement software that filters out all UPD traffic, which is what mDNS is based on. Mostly commonly, on MacOS, Microsoft Content Filter
is the culprit.
In order for mDNS to function properly, ensure Micorosft Content Filter
is inactive.
mdatp system-extension network-filter disable
and hit enter.Microsoft Content Filter is disbaled when the output is “Success”.
Some organizations will re-enable the filter from time to time. If you repeatedly encounter app-id values missing, first check to see if the filter has been re-enabled before doing more extensive troubleshooting.