Migrating routeless apps to Diego doesn't disable health-check

[dmarmugi]

As described in https://github.com/cloudfoundry/diego-design-notes/blob/master/migrating-to-diego.md#running-route-less-applications-such-as-workers-and-schedulers, a DEA --no-route is analogous to a Diego --no-route --set-health-check none

cf-migrate-apps diego should disable the health check on routeless DEA applications

Thanks,
-David

[cf-gitbot]

We have created an issue in Pivotal Tracker to manage this:

https://www.pivotaltracker.com/story/show/131912253

The labels on this github issue will be updated when the story is started.

[dkoper]

@socalnic @ematpl Wdyt? Kinda relates to the discussion we had a while back about whether cf push --no-route should imply health-check none.

[emalm]

/cc @SocalNick

[SocalNick]

If the plugin can properly determine whether or not to add --set-health-check none, would make it more useful.

[emalm]

My principal concern would be that there could be a stopped app that when started and routable should have a port-based health-check, but when migrated doesn't happen to have a route mapped to it right then (maybe it's the inactive half of a blue-green deploy pair, and the blue-green deploy technique is being parsimonious with route mappings). The health-check inference on cf push has a similar issue. An appropriate mitigation for that concern might be for the migration commands to report warnings about all apps with their health-checks set to none, so that the migrator can more easily discover the subset of apps for which to verify that this is the correct health-check type.

Also, can we use the clearer process instead of none for this?

Thanks,
Eric

[dmarmugi]

Hi all, thanks for considering this.

I can offer that as a user, I wouldn't have had any problem with it dropping me into an interactive (yes/no/yes-to-all/no-to-all) mode to quickly set the appropriate health-checks per app/space. It's essentially what I had to end up doing.

Thanks,
-David

On Oct 7, 2016, at 20:37, Eric Malm [email protected] wrote:

My principal concern would be that there could be a stopped app that when started and routable should have a port-based health-check, but when migrated doesn't happen to have a route mapped to it right then (maybe it's the inactive half of a blue-green deploy pair, and the blue-green deploy technique is being parsimonious with route mappings). The health-check inference on cf push has a similar issue. An appropriate mitigation for that concern might be for the migration commands to report warnings about all apps with their health-checks set to none, so that the migrator can more easily discover the subset of apps for which to verify that this is the correct health-check type.

Also, can we use the clearer process instead of none for this?

Thanks,
Eric

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or mute the thread.

[dkoper]

We explored interactive prompts when previously enhancing this plugin, but gave up because of its technical complexity (the plugin itself runs in a separate process). We'll need a lot more demand before we'd want to revisit that.

So it seems the safest solution would be to display a warning for each app being migrated to Diego that has no mapped routes that health-check-type is set to none.
(Will look at the stories to support process for the cf CLI and app manifest separately):

WARNING: Assuming process health check (type "none") for app with no mapped routes; use `cf set-health-check` to change this: App my-app to Diego/DEA in space myspace / org myorg as diesk

[dkoper]

@dmarmugi We have released v1.2.3 of the plugin with the warning as described above.
Thanks for bringing it up!