[SC-23705] fix(sematext): deduplicate corrupted measurement name prefixes

[boratanrikulu]

Summary

• Fix Apache metrics corruption where field names contained 64+ repeated "apache." prefixes
• Add deduplicatePrefix() function to normalize measurement names before building lookup keys
• Add unit tests for the deduplication logic

Technical Details

Root Cause

When originalName arrives corrupted (e.g., "apache.apache.apache..."), the field lookup key at rename.go:220 becomes:

"apache.apache.apache...BusyWorkers"

This doesn't match fieldReplaces["apache.BusyWorkers"], so it falls back to ChangeNames() which returns the corrupted name.

Fix

The deduplicatePrefix() function normalizes corrupted measurement names before building lookup keys:

• "apache.apache.apache" → "apache"
• "nginx.nginx.nginx" → "nginx"

This ensures the lookup key is always "apache.BusyWorkers" which correctly maps to "workers.busy".

Test Plan

• Unit tests for deduplicatePrefix() covering normal inputs and edge cases
• Integration test with corrupted measurement name verifying correct field rename

[otisg]

When originalName arrives corrupted (e.g., "apache.apache.apache..."), ....

But why does it arrive corrupted? Where does it come from in this corrupt format? Isn't that what we should figure out and fix the problem at the source?

[boratanrikulu]

But why does it arrive corrupted? Where does it come from in this corrupt format? Isn't that what we should figure out and fix the problem at the source?

Yes valid questions but I couldn't find the source so far. If I understand correctly it could be related to re-try logic but I'm not total sure.

[boratanrikulu]

even with patch, still seeing weird names, will close this for now.