How to do performance monitoring in a microfrontend architecture

[mohitpatel-ghl]

I have been digging up on this for a while, get in contact with support team, looked into open issues as well (#9601).
And i know explicitly performance monitoring for each micro frontend wise is currently not supported. But i wanted to know what's the recommended way or some custom solution so that we can get performance monitoring on Sentry. Because it kind of issue for us or anyone with micro frontends that error tracking is working fine but for performance monitoring we need a separate tool possibly a RUM based tool.

@AbhiPrasad
What i have observed is that we actually don't need segregation for performance micro-frontend wise or project wise. If we can get a single unified view for performance for whole app that would work. Like other RUM tools are doing.

And this is not the issue just we are facing but other companies as well with similar architecture, i also had discussion with one of the company who were using Sentry for error they were but they had to integrate different tool for performance.

I & my team really loves sentry for its error part ,we hope similar experience on performance side as well as it is equally important

[AbhiPrasad]

@mohitpatel-ghl I recommend you initialize performance with a single sdk instance that records everything for the site. That'll also capture web vitals like LCP, CLS. I don't recommend each individual micro-frontend instance sending spans/metrics.

they had to integrate different tool for performance

What does this tool do? Does it flush LCP for every single microfrontend? This doesn't make too much sense to me, and it'll lead to larger costs as you end up paying more for duplicate data.

[mohitpatel-ghl]

Thanks @AbhiPrasad for prompt response.

I recommend you initialize performance with a single sdk instance that records everything for the site.

I guess by this you mean to only initialze in the Host app & not in remote apps of micro-frontend. If this is what you are saying then we have done exactly same.
We have initialized single SDK only in our host app & then all other remote apps have their own dsn defined in Sentry Webpack Plugin. And as per docs we are using transport & multiplexer to route errors to different projects from host app.

So the error part is working great each micro frontend have their errors segregated to respective project.

I don't recommend each individual micro-frontend instance sending spans/metrics.

So the question over here is how do we stop micro frontend or remote apps to stop sending spans/metrics but only error events.

[mohitpatel-ghl]

@AbhiPrasad can you help here or point to right person of contact.