Closed thierroboto closed 11 months ago
Which version of Sentry are you using?
With the current timeouts in place it shouldn't happen that your site is affected unless you are sending MANY events to Sentry in a single request. Also performance transactions are sent after the response is sent to the user so that also shouldn't affect it. Not saying it wasn't but curious to learn more about your environment to figure out if there are things we could improve still to prevent affecting applications when Sentry has downtime.
I do also want to mention that the best way to prevent Sentry from affecting your application is running Relay, in the docs it is mentioned as a way to improve response times (which it also does) but as an added benefit if Sentry is down your Relay instance is not and will accept events and even queue them until they can be forwarded to Sentry. See: https://docs.sentry.io/platforms/php/performance/#improve-response-time.
Thanks for the quick reply. We're using the following:
"sentry/sdk": "^3.4", "sentry/sentry": "^3.20.1",
Should I update?
I am unsure if i'm sending many requests at the same time, but i just noticed my sample rate was at 1.0. I just brought it down to 0.2, will that help in the meantime?
I'm looking into the relay right now. We're using both heroku and AWS to host our applications, which would you suggest to deploy relay? if AWS, which service?
Thank-you
Those versions are new enough to include the latest timeout changes. So no need to upgrade 👍
Just to clarify, you are using the Laravel SDK right? Or are you using just the PHP SDK?
The laravel sdk
I do not recommend lowering the sample_rate
. This could cause you to miss out on important errors.
A local Relay can be hosted on any x86 Server/VM, so either ECS using our docker image or an EC2 instance running the binary directly.
I don't have much experience with Heroku though, so I can't give you any recommendations for it.
Our official recommendation is to use a local Relay https://docs.sentry.io/platforms/php/guides/laravel/performance/.
Problem Statement
As you know, Sentry event ingestion went down a few times in the past week. When that happens, our website goes down with it, as all the requests start timing out.
Some kind of failsafe system that would deactivate sentry temporarily in these cases would be very appreciated.
Solution Brainstorm
Perhaps a query to the Sentry status page to see if everything is operational before trying to send the events to Sentry?