3 ms·

> I do think there are situations where support can't help, but those are usually situations where Heroku is probably not the right place to host a service. Th

> I do think there are situations where support can't help, but those are usually situations where Heroku is probably not the right place to host a service. This hasn't been my experience at all. Mysterious performance issues are met with "We can't help, just install NewRelic" (the issues don't exist when running the same code on other services). Billing questions have yielded delayed and cryptic responses. Table stakes customer support simply isn't being addressed. And I don't blame the support agents, I just don't think they have the tools or resources to help. By comparison, AWS has managed to answer almost all of my questions within one or two messages. Hell, even Amazon has admitted when there's a bug on their end. That simply doesn't happen at heroku anymore.

jbm 4y ago
I hate that I missed this response because it's a fair criticism. Yeah, there are definitely details Heroku didn't have; for example, we only had something like the past 500 (or was it 100?) lines from a customer's logs; you would need to have an add-on to retain them. There were no internal logs to consult about app performance; the New Relic install requests were probably to see if there is an issue related to the app. If the issue went deeper, we would work with the engineering team for the feature in question; but with the volume of tickets, we would have to categorically rule out an app issue first as 99% of issues happened while I was there. I think it is fair to say Heroku's support tried to embrace the "teach a man to fish" strategy, and that wouldn't work if the problem wasn't with the app. I can't really speak to the billing issues, but that team was tiny; two or three people. I could definitely see them getting overwhelmed, and I heard about the lack of detailed billing support. (Out of curiousity, were you using Heroku's docker product? I could definitely see that as having caused the sort of problems you mention as Heroku's docker platform was not vanilla docker.)

jbm 4y ago

I hate that I missed this response because it's a fair criticism. Yeah, there are definitely details Heroku didn't have; for example, we only had something like the past 500 (or was it 100?) lines from a customer's logs; you would need to have an add-on to retain them. There were no internal logs to consult about app performance; the New Relic install requests were probably to see if there is an issue related to the app. If the issue went deeper, we would work with the engineering team for the feature in question; but with the volume of tickets, we would have to categorically rule out an app issue first as 99% of issues happened while I was there. I think it is fair to say Heroku's support tried to embrace the "teach a man to fish" strategy, and that wouldn't work if the problem wasn't with the app. I can't really speak to the billing issues, but that team was tiny; two or three people. I could definitely see them getting overwhelmed, and I heard about the lack of detailed billing support. (Out of curiousity, were you using Heroku's docker product? I could definitely see that as having caused the sort of problems you mention as Heroku's docker platform was not vanilla docker.)