50c2fe2a1e
ci / lint (push) Successful in 1m19s
ci / unit (push) Failing after 1m2s
ci / integration (push) Has been skipped
ci / types (push) Successful in 1m37s
ci / security (push) Failing after 38s
ci / dockerfile (push) Successful in 14s
ci / image (api) (push) Has been skipped
ci / image (reconciler) (push) Has been skipped
ci / image (worker) (push) Has been skipped
ci / bump (push) Has been skipped
Complete working build of the system learn-python/ teaches. 164 tests, mypy --strict clean, domain coverage 99%.
16 lines
970 B
SQL
16 lines
970 B
SQL
-- 004_traceparent.sql — carry the trace across the queue.
|
|
--
|
|
-- A trace is a chain of span contexts. `POST /v1/instances` commits a row and returns;
|
|
-- a worker in another pod claims that row minutes later. Nothing propagates the context
|
|
-- across that gap, because the gap is a table. So the context rides in the table: the API
|
|
-- writes the W3C traceparent it is currently inside, and the worker extracts it at claim
|
|
-- and makes its span a child of the API's. Skip this and Tempo shows two unrelated traces
|
|
-- for one provision, which is worse than no tracing — it looks like it works.
|
|
--
|
|
-- Nullable, no default, no backfill: expand/contract done right. Old rows have no
|
|
-- traceparent and never will; a worker running the previous image ignores a column it has
|
|
-- never heard of. Migrate first, deploy second, and neither step needs the other to have
|
|
-- happened.
|
|
|
|
alter table tasks add column traceparent text; -- '00-<32 hex>-<16 hex>-01', null = untraced
|