-
Notifications
You must be signed in to change notification settings - Fork 164
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Dispatch count to resolution metadata #1343
Conversation
@willvedd would you mind adding this test as well? The gist just exercises direct userset dispatches and is a good test to have. It's a little easier to reason about the direct userset lookups and their dispatches. Side note for future discussion and work, the test case(s) in the gist are a good example of something we can optimize in our Check algorithm by doing direct "bulk" lookups first, and then dispatching residual subtrees. For example, when we have to evaluate the subtree for |
Co-authored-by: Jonathan Whitaker <jon.b.whitaker@gmail.com>
…into dispatch-metrics
Note: Only added dispatch count metrics for check for now. Listobjects will require a follow-up PR once this is approved. |
Co-authored-by: Jonathan Whitaker <jon.b.whitaker@gmail.com>
…into dispatch-metrics
Description
What makes a check query expensive in both resources and time? Ostensibly, a high number of datastore queries is probably the first answer. However, that isn't the full picture and there are other aspects of a check query that could affect the cost. For example, the depth of the relationships, the number of tuples, the number of dispatches, etc. We have decent intuitions but ultimately we need data to understand the relationships between those facets and the bottom-line performance.
This PR is a first-pass at trying to unearth some check resolution metrics through metadata so that they can be reported on and analyzed.
References
Review Checklist
main