Ansible

240 readers

1 users here now

# TODO

founded 2 years ago

MODERATORS

NegativeLookBehind@lemmy.world

How to parallelize execution on a single host? (self.ansible)

submitted 3 weeks ago by wasabi to c/ansible@lemmy.world

6 comments fedilink hide all child comments

I'm setting up an application using containers with ansible. I want to be able to set up the same application multiple times with a different set of variables. Is there a way how I could do this in parallel on a single host? I know I could deploy the same application n times on n different hosts, but what about n times on a single host? Is something like this possible? Doing it sequential obviously works, but it doesn't scale well.

you are viewing a single comment's thread
view the rest of the comments

[–] DasFaultier@sh.itjust.works 1 points 3 weeks ago (2 children)

I just had another idea: you could use the Mitogen execution strategy. It doesn't make your tasks run in parallel, but it takes away a lot of the overhead involved in running Ansible, so you might get some performance gains there. It's not running your tasks in parallel though...

For real parallel execution, you would probably need multiple hosts. I don't think what you're trying to do is even a part of Ansible on the conceptual level.

[–] wasabi 2 points 3 weeks ago* (last edited 3 weeks ago) (1 children)

Could one maybe use multiple hosts in the ansible inventory that point to the same host?

Maybe k8s ist better suited for this.

[–] DasFaultier@sh.itjust.works 1 points 3 weeks ago* (last edited 3 weeks ago)

Yes, you could setup multiple DNS aliases in your nameserver/router or in your /etc/hosts file that point to the same IP; DNS might round-robin through the IPs, but I've never tried this.

EDIT: You're right, the DNS aliases go in the Ansible inventory, no need for DNS round-robin. Short brainfart, sry. EDIT END.

I have near-zero experience with k8s (apart from a little bit of Minikube, which I hated), so I can't say anything about that option, but my impression is that it's massive overkill for almost all use cases, aka "you're not Google".