We noticed an error today on a widget saying “Cannot find model x in scopes”. I submitted a ticket via email so I could include internal info more easily, but I also have an open question regarding monitoring. We only found the issue when an internal user noticed it and reported it to us. How do we monitor and get alerts for errors like that? We’re mostly concerned about canvas dashboards since those are our customer-facing dashboards.
AML validation still succeeds so we can’t use that
The error doesn’t seem to be due to a code change on our part (we’re in a code freeze), so I’m not sure that the recently released Data Tests would help here since there’s no way to automate these yet, and we’d have to have already set up a test for what seems to be a pretty specific edge case
Jobs Monitoring doesn’t seem to show any failed jobs with a failed status, either in the list of jobs or in the jobs monitoring analytics dashboard
Is there any way to expose some kind of list of dashboard errors, particularly in a way that we can set up alerts on them? Or is there validation logic for catching errors like this that we could automatically run on a set schedule or something? (ex. if AML Validation covered this, then we could automatically run that at a set schedule just to check that everything is working as expected and set up our own alerting)
My colleague has already handled your support ticket. This is a bug from our side while optimizing the AQL engine. A patch will be released to fix the issue.
Currently, our AML validation can’t detect errors that occur at runtime. However, we are actively researching ways to enhance our error detection capabilities. While Data Test project has not yet been released, it is part of our long-term roadmap.
We truly appreciate you bringing this to our attention and want to assure you that we have taken note of your feedback.
Ah, sorry about making it seem like I was reporting the issue twice, I was mostly just explaining the context for this ask around monitoring.
Understood about AML validation not being able to detect errors that occur at run-time, but I wonder if eventually there could be a way to “run” the dashboards automatically to check for errors? (At least maybe the canvas dashboards)
Or even potentially just some way to monitor for errors that have already happened. Ex. If a customer views an embedded dashboard and gets an error, I would LOVE to have some alerting set up so that we see the error without waiting for them to report it back to our internal support team.
Thanks Tri, appreciate the suggestions, we’ll look into them! Question on preloading a dashboard via the API - is there an easy way to get the error message if there is one? Or would that only be visible if someone is actually looking at the dashboard manually?
Hi @anya.conti,
You can get the error messages on your side. I’ll clarify the process a little bit:
Submit the Preload Dashboard API with a dashboard ( bust_cache param should be true). Many dashboards should run with many API calls. Each API call will create a Job run and return a Job ID.
Wait for each Job run (via Job Result API) until it’s done to get the dashboard result and see if it ran successfully or failed. You can take this code snippet as an example.
