Conversation
docs/10-clusters.md
Outdated
| ### Run a command on a specific node in the cluster | ||
| To run a command on all nodes in the cluster, run: | ||
| ```sh | ||
| aztk spark cluster node-run --id <your_cluster_id> --node-id <your_node_id> "<command>" |
There was a problem hiding this comment.
exec or node-exec? Stays more in line with docker exec
There was a problem hiding this comment.
There is already a cluster run so I figured cluster node-run made the most sense here.
|
|
||
|
|
||
| def print_execute_result(node_id, result): | ||
| print("-" * (len(node_id) + 6)) |
There was a problem hiding this comment.
should use the log here i think
aztk/client.py
Outdated
|
|
||
| def __node_run(self, cluster_id, node_id, command, internal, container_name=None, timeout=None): | ||
| pool, nodes = self.__get_pool_details(cluster_id) | ||
| node = [node for node in nodes if node.id == node_id] |
There was a problem hiding this comment.
would use next(node for node in nodes if node.id == node_id) here instead of this and the if below https://stackoverflow.com/questions/9542738/python-find-in-list
aztk/client.py
Outdated
| ssh_key = RSA.generate(2048) | ||
| ssh_pub_key = ssh_key.publickey().exportKey('OpenSSH').decode('utf-8') | ||
| try: | ||
| self.__create_user_on_node('aztk', pool.id, node.id, ssh_pub_key) |
There was a problem hiding this comment.
shouldn't this user have some kind of random id so it doesn't conflict
There was a problem hiding this comment.
maybe, but if the user already exists, then it is just deleted and recreated. So the only potential issue is if someone adds aztk as a user, and then runs a command. In that case, their created user would be deleted, so adding a randomly generated user id would probably be best here.
There was a problem hiding this comment.
or running node-run twice at the same time. Not sure what deleting the user while being login does
There was a problem hiding this comment.
I believe nothing would happen while the session is open. However, still, making the user id generated would solve that.
|
On second thought, for this PR, maybe the CLI should just be |
|
Thought this, the problem is the same we talked about earlier where you might make a mistake and run on all nodes instead of a single one. Is it however a big deal? |
aztk/client.py
Outdated
|
|
||
| import azure.batch.models as batch_models | ||
| import azure.batch.models.batch_error as batch_error | ||
| from azure.batch.models import batch_error as batch_error |
There was a problem hiding this comment.
is this import the same as above?
| log.info("%s\n", result) | ||
| else: | ||
| for line in result: | ||
| print(line) |
2 participants
Fix #571