Exec into the containers¶

If you want to examine the contents of a container “in place” using remote access to it, you can use the kubectl exec command. It allows you to run any command or just open an interactive shell session in the container. Of course, you can have shell access to the container only if container supports it and has a “Running” state.

In the following examples we will access the container mongod of the my-cluster-name-rs0-0 Pod.

Run date command:
```
$ kubectl exec -ti my-cluster-name-rs0-0 -c mongod -- date
```
Expected output
```
Thu Nov 24 10:01:17 UTC 2022
```
You will see an error if the command is not present in a container. For example, trying to run the time command, which is not present in the container, by executing kubectl exec -ti my-cluster-name-rs0-0 -c mongod -- time would show the following result:
```
OCI runtime exec failed: exec failed: unable to start container process: exec: "time": executable file not found in $PATH: unknown command terminated with exit code 126
```

Print /var/log/mongo/mongod.log file to a terminal:

$ kubectl exec -ti my-cluster-name-rs0-0 -c mongod -- cat /var/log/mongo/mongod.log

Similarly, opening an Interactive terminal, executing a pair of commands in the container, and exiting it may look as follows:

$ kubectl exec -ti my-cluster-name-rs0-0 -c mongod -- bash
[mongodb@my-cluster-name-rs0-0 db]$ cat /etc/hostname
my-cluster-name-rs0-0
[mongodb@my-cluster-name-rs0-0 db]$ ls /var/log/mongo/mongod.log
/var/log/mongo/mongod.log
[mongodb@my-cluster-name-rs0-0 db]$ exit
exit
$

Avoid the restart-on-fail loop for Percona Server for MongoDB containers¶

The restart-on-fail loop takes place when the container entry point fails (e.g. mongod crashes). In such a situation, Pod is continuously restarting. Continuous restarts prevent to get console access to the container, and so a special approach is needed to make fixes.

You can prevent such infinite boot loop by putting the Percona Server for MongoDB containers into the “infinite sleep” without starting mongod. This behavior of the container entry point is triggered by the presence of the /data/db/sleep-forever file. The feature is available for both replica set and confg server Pods.

For example, you can do it for the mongod container of an appropriate Percona Server for MongoDB Pod as follows:

$ kubectl exec -it my-cluster-name-cfg-0 -c mongod -- sh -c 'touch /data/db/sleep-forever'

If mongod container can’t start, you can use backup-agent container instead:

$ kubectl exec -it my-cluster-name-cfg-0 -c backup-agent -- sh -c 'touch /data/db/sleep-forever'

The instance will restart automatically and run in its usual way as soon as you remove this file (you can do it with a command similar to the one you have used to create the file, just substitute touch to rm in it).

Get expert help¶

If you need assistance, visit the community forum for comprehensive and free database knowledge, or contact our Percona Database Experts for professional support and services. Join K8S Squad to benefit from early access to features and “ask me anything” sessions with the Experts.

Community Forum Get a Percona Expert Join K8S Squad

Last update: 2024-04-26