Hosted proxy reverting to first revision a few hours after new revision deployment

[#1]

Hey all - experiencing an issue with hosted targets that I have run out of debug options for. I deployed an API proxy hosted target revision yesterday around 6pm EST. I tested the single endpoint in the proxy and found it worked without issue. This morning, around 10am EST (16 hours after revision creation), I tested the single endpoint again with the same script and found that I was receiving a response from the initial "Hello World" revision, not the revision I deployed the previous day.

My best guess is that with zero load on the proxy over night it scaled down to zero instances, and when I tested this morning it scaled back up to a single instance to handle the call, but instead of deploying the latest revision's image it deployed the first revision's deployment.

This is at least the second time I have seen this behavior with the same proxy.

I don't see anything in the runtime or build logs that suggest any issues, though my latest attempt to fetch the logs errored out with "Error fetching Category Invalid status code from backend deployment service (403): Token has expired or not active yet." I have logged out and back in and am still getting the same error, so that is curious.

Has anyone seen this behavior before, with either a hosted target or the traditional proxy service with the Java runtime? If so, have you resolved it or did it resolve itself?

I recognize the hosted targets product is in beta so there may be some issues still to work out. Perhaps this is an issue in the product that needs reporting.

Thank you much.

[bowenjosh]

Hey Jarret,

I work on the Hosted Targets team, thanks for trying out our beta.

I can't recall seeing the issue where an older revision is spun up before, though you are correct that the proxy will scale down to 0 instances. Could you provide me with some info like your org/env/proxy name so I can try to debug this(since this is reliant on a scale down it would be slow to reproduce on my end).

With regards to the logging issue, I've just noticed that this morning and am starting to look into it now. I'll post an update to this question when I know more.

Regards,

Josh

[#3]

Thanks for the quick response Josh, apologies for the super slow one on my part. Management here has asked me to avoid identifying the organization and the proxy in a public setting if it's at all avoidable. Is it possible to move the discussion to email or into a support ticket? Perhaps the best route is for me to open a support ticket (I didn't appear to have the ability to do so yesterday, but another at my organization says they have the ability to do so for me if needed).

For what it's worth I can confirm that I'm still seeing the behavior - I have avoided doing any other deployments in the event that it's better for debugging purposes to have it in its current state, especially since time appears to be a factor as you noted.

[sidd-harth]

Yes, open a support ticket & if you want to share some confidential info here in the community with Apigee folks, use the refer tab.