Notifications Improvement: Send error / warning notifications to API endpoint
AnsweredAbility to send error and warning notifications to an API endpoint so that they can be used with various paging tools (i.e., opsgenie, victorops, pagerduty). Also would like ability to only send errors / warnings so that notifications aren't sent when a initial load finishes or when Fivetran sends general notifications.
-
Official comment
Hi David Kincaid! We've officially launched the ability to send notifications to any email address. You can read up on how to use this in our documentation. Opsgenie, and other messaging/incident management systems, accept email based messages so this should allow for better integration with these systems. We've also documented steps to integrate emails with these systems as well.
-
Thanks David. Good news is we're currently working on a beta release that would enable you to send error (task) emails for all or a subset of your connectors to any email address. I don't have a timeline on this but would love to have you test it out once we're at that point.
-
Hi David- Thank you for submitting this feature request. We are currently looking at making some major improvements to our alert and notification system in the first half of 2021. One piece of functionality we're exploring is the ability to link Fivetran's alerts to different paging tools. If you could answer a few questions, that would be really helpful as we explore this further:
- What tool or tools do you use today?
- Would you want all Connector errors to generate a page, or a subset of your Connectors?
- Would leveraging an "email based" integration, such as this, work for your use-case over a specific API? Assuming we clean up the notifications to work well with email based alerting.
-
Hi David- Following up here and hoping you could answer the questions in my earlier comment. Thanks!
-
- What tool or tools do you use today? OpsGenie with an email integration
- Would you want all Connector errors to generate a page, or a subset of your Connectors? Subset of connectors. Some of our connectors are more critical than others so having the ability to configure which specific connectors generate pages seems to allow the most flexibility.
- Would leveraging an "email based" integration, such as this, work for your use-case over a specific API? Assuming we clean up the notifications to work well with email based alerting. This is what we use today. The challenge is that all Fivetran messages go through the email and not just errors / items needing to be actioned.
-
Hi FiveTran, is there any updates to this on the above? A callback to a webhook url would be great if there's an error or if a resync occurs. It's been 3 years and notifications can only be sent via Email only. Other tools like AirByte released with webhook url notifications when Sync failures are detected
-
Hi Lukey,
There's a lot of new functionality you can take advantage of, including webhooks. To track failures, you should be using the sync_end event.
You can also integrate Fivetran with external log services or send notifications to third party tools like Slack.
Let me know if you had something else in mind, but I think that should cover your use case.
-
Hi Lyndsey,
I've looked into the webhooks on the link you sent. Why is it that we have to setup a Python/Flask or NodeJS/Express application and host it somewhere ourselves?
In comparison Airbyte has a place where you can put a webhook url in their notification settings. Any sync failure sends an event to that webhook url. You can even press a button to send a test event to that webhook url.
Also we're looking into setting up notifications getting sent to third party tool (ServiceNow) however I've set up a connector that fails every 5 minutes. I've only received 1 email (the initial failure) but not from subsequent failures. I've opened up a support ticket for this. Could you explain this too?
-
Hey Lukey,
In regards to webhooks, you don't have to setup a flask/node project to receive webhooks, it's just sample code to show how you might receive/parse them – but there do have to be some endpoint setup to receive them. Our buttons are just endpoints vs a UI.
In regards to the notifications, yes, we intentionally only send one email for each new error (see "New errors" in our Docs). We don't want to spam customers with emails every time the sync fails, if it is failing with the same error. In your case, I'm curious if you would want to create multiple tickets in ServiceNow? Why?
If you want real-time information related to failures, webhooks would be a great option, or if you have a logging service, you can also set up alerting for each failure by using the sync_end event.
Lyndsey
-
Hi Lyndsey,
Thanks for your response. I see, but your official documentation suggests the flask/node project is required to setup the endpoint. If we don't have to what would you recommend?
I understand that you wouldn't want to spam customers every time the connector fails. However, we would not be creating a new ticket in ServiceNow, it would instead append to an existing incident. Using this we can do more such as tracking how long it took for the connector to be resolved. How many syncs failed before it was resolved etc.
Thanks,
Lukey -
Hi Lukey,
The webhooks question likely requires a deeper conversation, however, on the notifications I can suggest one thing – if you want to track how long it took for the error to be resolved from a Fivetran perspective, we also send an error resolved email. If you can append this to the existing incident, you should have the difference between when the error first occurred and the time of error resolution.
For how many syncs before it was resolved, I'd typically recommend using logs and count the number of sync events. That will likely need to happen outside of ServiceNow as we don't send notifications for each sync. Aside from external log services, this data is available via our Fivetran Platform Connector.
Lyndsey
Please sign in to leave a comment.
Comments
11 comments