Dynamic Sampling Launch #41017
Replies: 1 comment
-
We collect CI information about all of our GitHub actions in Sentry via https://github.com/getsentry/sentry-github-actions-app - a GitHub application that @armenzg built. With this app, we instrument all of our GitHub actions and send that data to our performance product. A CI job run -> a Sentry transaction. See the project that contains this data: https://sentry.io/organizations/sentry/projects/sentry-github-action-ci-data/?project=5899451. To get further insights, this transaction is tagged with a variety of tags we use for filtering. In particular, With dynamic sampling and the sampling volume at Sentry, there is a significant difference between processed and indexed transactions. This is problematic for the CI project since we require filtering the data by specific tags to do meaningful analysis of the data. Since the Unfortunately by adding this For example, we have a widget that we use to track the most common CI failures on the master branch. Right now in the past 30 days, only 3 indexed transactions were ingested that have the This means we cannot effectively use dashboards with the current state of dynamic sampling on our CI data. CI data is also low volume in nature, so even the indexed transactions are low volume since they get drowned out by higher volume repos (same project, but different sources for each transaction). To fix this, I would like to either a) Have the ability to turn off dynamic sampling for a certain project - I am willing to pay extra to get as much fidelity as possible into these transactions |
Beta Was this translation helpful? Give feedback.
-
Just because you don’t record a problem, that doesn’t mean it didn’t happen.
This week we launched Dynamic Sampling - available to all new customers, and opt-in for existing ones. Dynamic Sampling is rebuilding on our original approach to data volume scale with a focus on maximizing the value out of your performance data - automatically.
What’s new in this approach?
We started a closed beta for Dynamic Sampling back in July with a strong hypothesis that our customers know best what data is most valuable to them. This made us focus on providing fine controls to target specific transactions and configure sampling rules directly in the Sentry product.
Talking to our early adopters, we found that our users loved the idea, but were still facing the same old questions: “what sample rate should I choose?”, “how do I make sure I get the most value?”, and many wanted Sentry to be more opinionated on that. Big shoutout to all of you who provided feedback and helped us understand how to evolve this 🙂
That’s why we’ve decided to remove the micro-management of sampling rules entirely. Instead, we’ll make intelligent decisions to automatically capture the data we believe is most valuable for most customers and the right amount based on your traffic. At the same time, we wanted to keep the ability for customers to influence how data is sampled, so we’re providing a mechanism to express priorities, so they can get the most value out of the transactions we store and filter out the noise.
We’ll be adding more capabilities in the future, but we’re starting out with:
Opting In
We’re rolling this out iteratively and today it’s only live for new customers. Existing customers will be getting this functionality soon, but if you’re excited to get started, we might be able to opt you in today — just contact your account rep or [email protected].
We want to hear what you think!
We really want to get your feedback on this:
Please post below to let us know. If you have any questions or feedback you want to give privately, you can also email me at [email protected]. Thanks a lot!
Beta Was this translation helpful? Give feedback.
All reactions