How to edit a kubernetes resource from a shell script - shell

I went through the documentation to edit kubernetes resource using kubectl edit command. Once I execute the command, the file in YAML-format is opened in the editor where I can change the values as per requirement and save it. I am trying to execute these steps by means of sed. How can the following steps be achieved?
Execute kubectl edit for a deployment resource
Set a value from true to false (using sed)
Save the changes
I tried to achieve this in the following way :
$ kubectl edit deployment tiller-deploy -n kube-system | \
sed -i "s/\(automountServiceAccountToken:.*$\)/automountServiceAccountToken: true/g"`

Your command is missing a backtick. But even though you put it there, it won't work. The reason is because when you do kubectl edit ..., it edits the file on vim. I am not sure sed would work on vim though. Even though if it does, the output goes to a file, so you get the Vim: Warning: Output is not to a terminal error, which I don't know how to solve.
I would recommend you to get the file and save it. Replace the desired parameters and run it again:
kubectl get deploy tiller-deploy -n kube-system -o yaml > tiller.yaml && sed -i "s/automountServiceAccountToken:.*$/automountServiceAccountToken: true/g" tiller.yaml && kubectl replace -f tiller.yaml
I tried the command above and it worked.
Note: no need to add -n kube-system as the yaml file already contains the namespace.

I just found a less convoluted way of doing this:
KUBE_EDITOR="sed -i s/SOMETHING TO CHANGE/CHANGED/g" kubectl edit resource -n your-ns

I automate through piping the commands through sed command without creating a temporary file. Take the below example, where I am replacing nameserver 8.8.8.8 with 1.1.1.1
$ kubectl -n kube-system get configmap/kube-dns -o yaml | sed "s/8.8.8.8/1.1.1.1/" | kubectl replace -f -

Thanks, #suren for giving what I really looking for, but you don't need to save it in a file. you can directly do kubectl replace using pipe operations
kubectl get deploy test-deploy -o yaml | sed "s/find/replace/g" | kubectl replace -f -

An easy way to do this, just use kubectl-patch instead of sed.
$ kubectl patch deployment tiller-deploy -n kube-system --patch '{"map": {"to": {"the": {"key": {"automountServiceAccountToken": "true"}}}}}'

I don't know kubectl but doc seems to explain that it extract data, edit from an editor than send back, not sure sed pipe work in this case
if piping wokrs
Don't use -i, you don't change a file in a pipe
kubectl edit deployment tiller-deploy -n kube-system | \
sed 's/automountServiceAccountToken:.*$/automountServiceAccountToken: true/g'
if editing a file (and using group in sed)
kubectl edit deployment tiller-deploy -n kube-system > YourCOnfigFile && \
sed -i 's/\(automountServiceAccountToken:\).*$/\1 true/g' YourConfigFile \
&& Some kubectl to send back YourConfigFile

Related

Error from server (NotFound): pods "\nmongo-client-79667cc85d-tsg72" not found

I'm trying to make a backup from Mongo / K8S with this script
export POD=$(kubectl get pods -l app=mongo-client -o custom-columns=:metadata.name -n espace-client)
kubectl exec "$POD" sh -c 'mongodump --archive' > ~/backup/mongo/$(date +%F).db.dump
I get this error:
Error from server (NotFound): pods "\nmongo-client-79667cc85d-tsg72" not found
When I check the pods, I can see mongo-client-79667cc85d-tsg72
When I put the name without variable, it works well, so it might be because of initial \n. How can I avoid it ?
How can I remove it from name ?
Your kubectl get pods command is constrained with a namespace selector -n espace-client. Your kubectl exec command also needs the namespace flag.
The output of your kubectl get pods command has a newline before the pod name because the first line of the output is the column header (which is empty in your case).
To prevent this and get only the name as output, you can suppress the column headers with the --no-headers flag:
kubectl get pods -l app=mongo-client -o custom-columns=:metadata.name -n espace-client --no-headers

View kubectl logs page by page on bash terminal

Team,
On bash I run below to view logs but when they are huge the terminal all is all occupied dumping them out and i can't stop it for minutes.
Is there a way i can execute kubectl on bash in a a way that it shows logs only till shell screen size and then i use return key or spacebar to see more ? just like journalctl? but i want to achieve this when using kubectl.
kubectl logs test-pod -n namespace-test
Above displans 10K lines at a time which i don't want. Neither I want to exec on to the pod and see log file physically. any hints? or is there something like which displays last 100lines? or first with xargs?
There are a few ways to obtain that.
1) less
kubectl logs test-pod -n namespace-test | less
This command will allow you to read logs page by page from top to the bottom. You can use arrows to go up or down.
2) --tail=
kubectl logs test-pod -n namespace-test --tail=500
Will display last 500 rows of logs
3) pipeline with grep
kubectl logs test-pod -n namespace-test | grep <some_phrase>
It will allow you to find logs with given phrases. Please note that you can join commands via pipeline i.e
kubectl logs test-pod -n namespace-test --tail=500 | grep <some_phrase>
4) --since=
kubectl logs test-pod -n namespace-test --since=60
It display logs from last 60 seconds
5) --since-time=''
Similar to previous one. It will display logs from provided time. Format in this command is
'YYYY-MM-DDTHH:MM:ssZ'
kubectl logs test-pod -n namespace-test --since-time='2019-04-23T12:00:00Z'
You could hear about more option like but it is old format. You can scroll only from top to the bottom, you cannot go up.
kubectl logs test-pod -n namespace-test | more
You can do:
kubectl logs --tail=100 test-pod -n namespace-test

How to forward psql shell from kubectl

I'm trying to make my life easier and am coding a bash script. One of these allows me to kube into a pod with postgres access, get the credentials I need, and run the interactive psql shell.
However, upon running
kubectl <flags> exec $podname -- bash -c ' get_credentials && psql <psql args> -i -t
the terminal hangs.
I can't directly connect to the database, and the process to get the credentials is kinda cumbersome. Is there some bash concept I'm not understanding?
kubectl <flags> exec $podname
That exec is missing its -i and -t for --stdin=true and --tty=true to describe to kubernetes that you wish your terminal and the remote terminal to be associated with one another:
kubectl exec -it $podname -- etc etc
If you are intending the -i and -t present at the end of your cited example above to be passed to exec, be aware that the double dashes explicitly switch off argument parsing from kubectl, so there is no way it will see them

Exec sed command to a docker container

I'm trying to change a config file that is inside a docker container.
docker exec container_name sed -ire '/URL_BASE = /c\api.myapiurl' tmp/config.ini
Executing this sed command locally works just fine, but when I try to execute this in the container I receive the following error message.
sed: cannot rename tmp/config.ini: Operation not permitted
What I need to do is replace the 'URL_BASE =' from the 'config.ini' before deploy the container to my server.
I don't know why the sed command is trying to rename the file when its not suppose to.
Any ideas?
What I've tried
I tried to execute with the --privileged flag, but didn't worked. I tried to change the file permissions with chmod but I couldn't for the same reason of permission.
docker exec --privileged container_name sed -ire '/URL_BASE = /c\api.myapiurl' tmp/config.ini
Result: sed: cannot rename tmp/config.ini: Operation not permitted
Chmod
docker exec --privileged container_name chmod 755 tmp/config.ini
Result: chmod: changing permissions of 'tmp/config.ini': Operation not permitted
I also have tried execute with sudo before docker but didn't work either.
Nehal is absolutely right, sed works creating a local file so you just need a different approach, which is commonly used on Linux: heredocs.
Taking just the first lines from the documentation, a here document is a special-purpose code block. It uses a form of I/O redirection to feed a command list to an interactive program.
It can help us with docker exec as follows:
docker exec -i container_name bash <<EOF
sed -ire '/URL_BASE = /c\api.myapiurl' /tmp/config.ini
grep URL_BASE /tmp/config.ini
# any other command you like
EOF
Be aware of the -t, which is commonly used running bash, because it allocates a pseudo-TTY, and we don't really need that.
Also, to be safe always use absolute paths like /tmp/config.ini.
docker exec -i <container name> sed -i 's/xxx/${yyy}/g' path/filename.yaml
This is working for me.

Bash grep IP from command output

I'm working with fleetctl and kubectl and would like to extract an IP from kubectl get pod app-etcd:
POD IP CONTAINER(S) IMAGE(S) HOST LABELS STATUS CREATED MESSAGE
app-etcd 10.10.0.1 k8s-socius-node-1/100.100.100.100 name=app-etcd Running 3 days
app-etcd xyz/etcd-discovery Running 3 days
The closest I got to get the IP address is:
kubectl get pod app-etcd | grep -Eo '(25[0-5]|2[0-4][0-9]|[01]?[0-9][0-9]?)\.(25[0-5]|2[0-4][0-9]|[01]?[0-9][0-9]?)\.(25[0-5]|2[0-4][0-9]|[01]?[0-9][0-9]?)\.(25[0-5]|2[0-4][0-9]|[01]?[0-9][0-9]?)'
But this gives me both IP addresses (10.10.0.1 and 100.100.100.100); I only want/need the first one to run sed over a config file afterwards.
How do I get only the first address to store it in a variable for further processing?
What if you use awk to get the second column output like
kubectl get pod app-etcd | awk '{print $2}'
kubectl offers json output with --output / -o:
kubectl get -o json pod app-etcd | jsawk 'return this.status.podIP'

Resources