Nomad: Zero Downtime Rolling Deployment using Health Checks

[robsonpeixoto]

I'd like to know if the nomad use the check to grantee a zero downtime upgrade.

[dadgar]

It currently does not but it is something on our roadmap!

[c4milo]

?

[robsonpeixoto]

Good question @c4milo

[skozin]

Yup, I'm also interested why this was closed. Was it removed from the Nomad roadmap?

This feature is very important for our use case, and its lack is one of the few things that prevent us from fully migrating to Nomad in our production clusters.

Currently we use Ansible to perform rolling deploys, while ensuring that the next instance of a service has transitioned into a healthy state before shutting down the previous one, but we'd love to replace this custom in-house logic with the one baked into Nomad.

[dadgar]

Was closed as it was a question but can rename and reopen

[skozin]

@dadgar, makes sense, thanks! :)

[igrayson]

We're eagerly awaiting this feature, so we can give nomad a try in production. For now, I don't believe we can risk the availability loss of using a deployment tool that doesn't integrate health checks to stop bad deployments (or the cost of building additional tooling to add that safety).

[igrayson]

Hi team,

Does this have any scheduling that we can learn about? Unless there's some recommended way to achieve zero-downtime deployments using nomad, this issue prevents us from considering nomad as a deployment tool.

[diptanu]

@igrayson We might start working on it soon!

[igrayson]

@diptanu Great!

I'm still a little confused. The current nomad docs themselves make reference to "zero downtime" deployments.

Is there any recommended approach to achieving zero downtime deployments, using nomad?

[dadgar]

Hey,

Automated zero-down time deploys require using health checks but the
covered blue/green section in the docs allows zero down time deploys:
https://www.nomadproject.io/docs/jobops/updating.html

Thanks,
Alex

[dadgar]

Closing in favor of #317