- [Russ] When you sit down to write an automated deployment, it very, very rarely works first time. When I'm doing this, I generally test some deployment commands manually. I get confident the deploy is
going to work how I planned. I start small, and gradually
add to the build scripts. That way, I can isolate debugging down to just the changes I
made in my small updates. When I do hit an error in
a CodeDeploy deployment, where do I start with my debugging? Number one, I can see all
the deployment details via the console. I'm looking at a failed deployment
in the console right now, and I'm seeing the message: "some instances in your deployment group are experiencing problems." I can further drill down into the instance details down here, and I see the deploy failed at the very first event, ApplicationStop. I'm thinking that I don't have
the CodeDeploy agent running on the instance. We need the CodeDeploy agent
to be installed in the instance for any deployments to work. Maybe you are installing this with your user data
scripts on your instances, or maybe you use AWS Systems
Manager to do this for you. I'll include some links
in the course readings with details on how to
install the CodeDeploy agent. I've connected to the
instance via Session Manager. I can check the status
of the CodeDeploy agent with this command: sudo service codedeploy-agent status. And that is telling me the
service cannot be found. That's a sign I need to
get the agent installed. What other types of errors
can occur in our deployments? I have fixed my CodeDeploy
agent on my instance, I've re attempted the deployment, and I see a failure. Again, in the console, I can see the details of the deployment. If I view the instance details down here, I'm being told: "scripts/start_server run as
root failed with exit code 5." If I view more on this, there
is the output of the script. It looks like I have a typo in there. CodeDeploy makes the
log tail available to me via the console and via the APIs. What if all my scripts ran, nothing returned an
error back to CodeDeploy, but there is something wrong
in my deployment logic, and I don't see my application working? Now, I would look at the
logs on the instance itself. Again, in a Session Manager
connection to my instance, I can see the deployment
log at this location: (typing) opt/codedeploy-agent/deployment-root/deployment-logs and there is a
codedeploy-agent-deployments.log. At the very end here, I can see the same error that
I was seeing in the console. I have a typo there in
"service httpd start." There is also a deployment
archive on the instance at this location. It is in
opt/codedeploy-agent/deployment-root. In here is the ID of
my deployment group ID, (typing) and under that is all
of the deployment IDs that have occurred on this instance. Let's choose one of them. And if I look here, I will
see the deployment archive that was downloaded to my instance, and a scripts.log for
this particular deploy. The script logs are going
to capture the output of my commands in my deployment scripts, and any logging I've included. This should point me to some clues of why my deployment isn't going to plan. So, where are you going to start your deployment detective work? We can see a lot of
information in the console. Do you need some extra debugging details? We can always retrieve CodeDeploy logs from our deployment instances.