fix(analytics): wait for logflare before starting vector

[okedeji]

What kind of change does this PR introduce?

Bug fix

What is the current behavior?

Vector starts before Logflare is ready to accept requests. During startup, Logflare returns 401 Unauthorized while its auth is still initializing. Vector treats 401 as a non-retriable client error (hardcoded in Vector's HTTP retry logic — only 408, 429, and 5xx are retried) and permanently drops those log batches. This results in missing startup logs in Studio and noisy error output in Vector's container logs.

What is the new behavior?

Vector's entrypoint now polls Logflare's health endpoint (/health) every 2 seconds before starting. Vector only launches once Logflare is confirmed ready, eliminating the 401 errors and ensuring all logs are delivered from the moment Vector starts collecting.

Tested locally — before: 37 x 401 errors, after: 0.

Additional context

Vector's HTTP sink retry logic is hardcoded and not configurable — 401 will never be retried regardless of retry_max_duration_secs. The wait loop uses wget which is already available in the Vector Alpine image (used by its own healthcheck).

[aantti]

@singh-inder, @luizfelmach fyi

[coveralls]

Pull Request Test Coverage Report for Build 23560482012

Details

• 3 of 3 (100.0%) changed or added relevant lines in 1 file are covered.
• 5 unchanged lines in 1 file lost coverage.
• Overall coverage decreased (-0.02%) to 63.165%

---

Files with Coverage Reduction	New Missed Lines	%
internal/utils/git.go	5	57.14%

Totals
Change from base Build 23560420141:	-0.02%
Covered Lines:	9193
Relevant Lines:	14554

---

💛 - Coveralls