Dynamic sampleRate for errors without re-deployment

[IgnusG]

A little bit of background, we expect a large amount of events to be sent to sentry during certain events or by accident (happened before, might happen again).

We would like not to collect all of these events as saving all quickly eats up the error quota and kills off other projects. There are some features that are available to prevent this like spike protection (unable to configure what constitutes a spike, entirely possible sentry does not trigger a spike and events are siphoned from the quota), spending allocations (only available on business+ plans) or rate limiting (also only available on business+ plans).

Ideally it would be possible to sample the errors based on their uniqueness like sentry used to do apparently according to this blog. But alas right now as a solution sampleRate sounds good enough. Simply turn it down to something like 0.1 or even 0 for periods with spikes and then re-enable it once the spike has passed. The problem is the documentation mentions this needs to be done through redeployment. As far as I could tell it's not possible to adjust the sample rate after the client has been initialized without re-initializing the client (I'm not sure what consequences this has?). We are considering doing this to bring the sample rate down immediately to preserve the quota and bring it back up once the spike event subsides. Redeploying the application is a last resort option if we can't find a workaround.

Does anyone have any tips on this or how/if they approached a similar situation?

I've also found documentation and blog posts from last and early this year about dynamic sampling (DS) but this seems to only be available for traces so far and the documentation is very difficult to parse. I'm still not sure if this is only available to high volume clients or higher tier plans or all plans behind a beta toggle since the blog mentions all tiers but the documentation often refers to Relay which is only available on business+ and I found it hard to understand how to activate it if it's even already released.

[lforst]

Hi, currently we don't have anything like that out of the box in the product.

If you wanna roll it yourself, you could always add custom sampling decisions in either of tracesSampler, beforeSendTransaction, or beforeSend in which you sample based on data your app fetched from your server.

[IgnusG]

Resolved using a suggestion made by @mydea to add a global event processor and adjust which events are sent using a "custom" sampler implementation