rusackas commented on PR #45601: URL: https://github.com/apache/superset/pull/45601#issuecomment-6333598850
Numbers are in. Attempt 1 on 48ae275 was the miss (loads as before, then dumps and saves, 24 MB compressed); attempt 2 is the hit. Master and the sibling PRs from the same evening are the before column. | "Import test data" step | master / sibling PRs | this PR, miss | this PR, hit | |---|---|---|---| | playwright-tests (chromium) | 257 to 393 s | 389 s | 34 s | | playwright-tests (chromium, /app/prefix) | 284 to 400 s | 265 s | 30 s | | playwright-tests-gaq (chromium) | 334 to 434 s | 343 s | 34 s | | playwright-tests-gaq (chromium, /app/prefix) | 319 to 405 s | 372 s | 38 s | The 30-odd seconds on a hit are the 2 s cache restore, about 13 s of `pg_restore`, and the `db upgrade` and `superset init` safety nets. The two GAQ jobs, where the import was most of the job, went from 12 to 13 minutes down to 5m28s and 6m23s; the two full Playwright jobs from 14 to 17 minutes down to 11 to 12. With the key hashing all of `superset/**/*.py`, a backend PR pays the miss once on its first run and hits on every push after that; frontend-only PRs hit the cache master built. Across the four jobs that's roughly 20 minutes of runner time per E2E run saved whenever the key matches. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
