Skip to main content

Community

Connector Improvement: Salesforce

Answered

Please sign in to leave a comment.

Comments

3 comments

  • Official comment
    Fraser User

    We've heard customers with hypothetical concerns about quota consumption, but in practice is doesn't happen.  Have you actually had quota issues caused by Fivetran's connector?

    By design, Fivetran is a good citizen of your Salesforce API. The Salesforce connector automatically pauses when your *global* rolling API quota reaches 90% consumption. It will not query the API when the quota is low.

    We did a survey of Salesforce connectors doing historical syncs & found they can sustain ~1 request per second.  This means that, at most, the Salesforce connector can consume ~87,000 API calls per day.  When doing incremental syncs, it is performing change data capture of update and deletes and uses much fewer API calls.

    Thanks for your answer.

    Please keep in mind the following:

    • At the connector level it is not specified that by default Fivetran is checking the quota (first point) and having a hard-stop at 90% (second point). Recommendation would be that this is shown for the customers. Eventually you folks could expose the current quota and I would like to challenge which error you receive when you reached that quota, if it's clear enough to the user.

      For big companies where we have Salesforce as main part of the Techstack, stopping at 90% is not enough and can collapse the business. Sometimes reaching that threshold is not a direct effect of the connector, but if our company would reach 75% we would like to triage certain processes that connect to Salesforce and de-activate until our consumption goes to a more safe number. In our case BI can be paused for a working day until we have more available API requests, where other parts of the business can't stop.
    • The calculation is correct on max call per day, what is not clear for the user is that you guys only use one connection. That's something that perhaps can be clarified, on our eyes we could understand that you could implement this as 1 connection per table you want to sync. This could explain why the syncing process to SF is rather slow sometimes, since you guys only use one active connection.

      It's not that it's super bad but we are not well aware how internally works :)
  • Ray User

    Hi BI,

    This is really great and thoughtful feedback. Thank you! As we try to make data more and more easily accessible for our customers, it's always important to think critically about how we surface the technical details and configuration options without overwhelming users with common use cases while still empowering users with more demanding requirements. We'll take these concepts under consideration for new improvements to our documentation and setup steps.

    Best,

    Ray Harris

    Product Manager, Sales & Customer Success Analytics