feat(nuxt): Always add tracing meta tags #13273
Conversation
s1gr1d
requested review from
Lms24,
a team and
chargome
and removed request for
a team
packages/nuxt/src/runtime/utils.ts
Outdated
| head.push(`<meta name="baggage" content="${dynamicSamplingContext}"/>`); | ||
| } | ||
| const metaTags = getTraceMetaTags(); | ||
| head.push(metaTags); |
There was a problem hiding this comment.
l: Not sure if this is problematic but I see we did this differently before: If nuxt expects one tag per array item we should split the return value from getTraceMetaTags() at \n since we usually return two tags separated by a \n.
There was a problem hiding this comment.
I checked this and nuxt puts the html together in the correct way so this should be fine. But I would rather add a function parameter { options.asArray } to return it as array instead of adding \n and splitting it again afterwards.
There was a problem hiding this comment.
I created a PR for this: #13293
There was a problem hiding this comment.
As I closed the PR I just use the returned string now. This works with Nuxt as well and we don't have to split the string (slightly better performance).
| ].join('\n'); | ||
|
|
||
| // return value is mocked here as return values of `getTraceMetaTags` are tested separately (in @sentry/core) | ||
| (getTraceMetaTags as Mock).mockReturnValue(mockMetaTags); |
There was a problem hiding this comment.
I think that's fine, given the unit under test here is the Nuxt-specific addSentryTracingMetaTags function.
What we should do in the near future is to make sure our e2e tests cover distributed tracing behaviour. For example, we should check that
- the sever-side request transaction and the client-side pageload transaction is connected by the same trace. This will validate that the trace meta tag injection generally works.
- potential requests made on the client side to the server (thinking of API endpoint or data loading calls, whatever nuxt offers) are connected to the server side. This is probably easiest testable when making these calls during or directly after a navigation so that we can make sure a client-side root span exists.
I'm also working on a SvelteKit e2e test app that's specifically configured for tracing without performance. We could think about doing something similar for Nuxt but I wouldn't require it for the moment.
Making sure tracing without performance works.
ref (SvelteKit): #13231