feat(mep): fix header filter error

[onewland]

I mistakenly thought that the KafkaHeaderFilter was passing matches but it was actually dropping them, so we were dropping all MEP messages (which is fine, it's pre-prod).

This fixes the issue

[evanh]

Why wasn't this caught with an existing test?

[onewland]

Why wasn't this caught with an existing test?

If you look at the original PR #2929, it felt more like a config than a functionality change. Any test would basically be mimicking the tests in test_message_filters

[fpacifici]

Still regarding the test. Did we already create the end to end test for this dataset? That should have caught it

[codecov-commenter]

Codecov Report

Merging #2938 (e669605) into master (5d9755c) will decrease coverage by 0.05%.
The diff coverage is 100.00%.

❗ Current head e669605 differs from pull request most recent head a2b5d2b. Consider uploading reports for the commit a2b5d2b to get more accurate results

@@            Coverage Diff             @@
##           master    #2938      +/-   ##
==========================================
- Coverage   92.93%   92.87%   -0.06%     
==========================================
  Files         635      635              
  Lines       29069    29081      +12     
==========================================
- Hits        27016    27010       -6     
- Misses       2053     2071      +18     

Impacted Files	Coverage Δ
snuba/datasets/message_filters.py	95.83% <100.00%> (+1.23%)	⬆️
snuba/datasets/storages/generic_metrics.py	96.15% <100.00%> (ø)
tests/datasets/test_message_filters.py	100.00% <100.00%> (ø)
snuba/settings/settings_distributed.py	0.00% <0.00%> (-100.00%)	⬇️
snuba/settings/settings_test_distributed.py	0.00% <0.00%> (-100.00%)	⬇️
...ts/0010_groupedmessages_onpremise_compatibility.py	95.55% <0.00%> (-4.45%)	⬇️
snuba/optimize.py	87.50% <0.00%> (-3.75%)	⬇️
snuba/migrations/table_engines.py	95.50% <0.00%> (-3.38%)	⬇️
...nsactions_onpremise_fix_orderby_and_partitionby.py	81.33% <0.00%> (-2.67%)	⬇️
tests/migrations/test_runner_individual.py	91.20% <0.00%> (-2.20%)	⬇️
... and 1 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5d9755c...a2b5d2b. Read the comment docs.

[onewland]

Still regarding the test. Did we already create the end to end test for this dataset? That should have caught it

The way we write e2e tests (at least since I've joined) directly invokes process_message on the payload (which in production comes after the prefilter):

https://github.com/getsentry/snuba/blob/master/tests/test_generic_metrics_api.py#L114-L136

so we can't apply a header-based pre-filter with this mechanism. To do that I think we'd need to build an actual consumer /processing strategy factory (e.g. https://github.com/getsentry/snuba/blob/master/snuba/consumers/consumer_builder.py#L221-L265)