[HG] Update Error Message "Inspect Grafana Server Logs" to contact Support

[timlevett]

Overview

Users on Grafana Cloud continuously run into the messages "Please inspect Grafana Server Logs" when an error occurs in the Grafana, however these logs are not available to them. This is confusing for the user who then goes looking for the logs or who the Server Admin might be and will not find them. Eventually they might open a Support ticket but they are also likely to give up without further guidance.

Suggested fix

Since currently "only" Support has access to these logs, the UI messaging should direct the user to open a Support ticket that notes the time in UTC at which the error occurred and what the user was doing when it happened.

internal reference

[timlevett]

Dev notes, we may want to make this configurable in the config.ini since on prem it still makes sense to say "inspect the server logs"

[evictorero]

I have been digging a bit about this. There are several places in Grafana (HTML and go files) where we send these messages to the users. We would probably need to use text/template go package to send back one message or another depending on where Grafana is running. There is an ongoing effort about I18n and also on standarize our Error responses. I think we could leverage both initiatives in order to parameterize these particular messages depending on if they are running in a Hosted cloud or not and maybe it would be better to not invest much time making changes right now which would probably be discarded in a couple of months.

[sakjur]

I'd suggest not using templating for that and rather have an option, outside of the error message, that adds a prompt to the user that there is detailed information in the logs that they're likely to be interested in, and then make that prompt overridable. The new Error type from #47504 that Eze mentions would be able to attach that as part of the public payload (that is of no interest to the backend, afaict), and the frontend can then act on that.

[jalevin]

Reviewed 11/23. Deprioritized, but we should do this at some point.

[mjseaman]

Hey, this has cropped up again and looks like it's still causing considerable confusion and weekly tickets for Cloud customers. Worth prioritizing for FY24 Q2?

[timlevett]

We should tackle this. I think we could generalize the message as a first step and then take the more complicated solution later. This could be a good first task for one of the operator experience squad members newer to the code base. Or we could shift it to front end platform if that makes more sense. Either way, I agree we need to tackle this.

On Thursday, April 27, 2023, Mitch Seaman @.***> wrote:

Hey, this has cropped up again and looks like it's still causing considerable confusion and weekly tickets for Cloud customers. Worth prioritizing for FY24 Q2?

— Reply to this email directly, view it on GitHub https://github.com/grafana/grafana/issues/49441#issuecomment-1525146138, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA265UHJ5EXPSMFFOMX2TQ3XDIWTNANCNFSM5WW5WYLQ . You are receiving this because you authored the thread.Message ID: @.***>

-- Tim Levett Engineering Manager | Grafana Labs https://grafana.com/ @.***

[ddorman16]

just following along / prodding here a bit. This would be a great quality of life improvement for our end users (and is one of those just things a user would expect when using SaaS)

[mmandrus]

PR open here to make this error message configurable. Once that is merged, the SaaS squad will have to update the default config for every instance. That update will slowly roll out as instances are reset.

[mmandrus]

Unless there are any objections, I am going to use this in grafana cloud:

[some high level error] - please try again. If the issue persists please open a support ticket that notes the time in UTC and steps to reproduce

CC @mjseaman @ddorman16

[ddorman16]

thx @mmandrus

out of curiosity, is the last part necessary? Think could we shorten the last part to end after '...open a support ticket."?

[mmandrus]

@ddorman16 imo it's potentially useful info for debugging, especially given that the user won't see any other error details. Does our support ticket template happen to ask for an approximation of when the event occurred? Maybe it would make more sense to have that.

[ddorman16]

the support ticket template asks for What happened? What was expected to happen? Steps to reproduce the problem. And option to include a screenshot.

So maybe

[some high level error] - please try again. If the issue persists please open a support ticket that notes the time in UTC.

[mmandrus]

That sounds better to me 👍