nic-6443 commented on PR #14055: URL: https://github.com/apache/apisix/pull/14055#issuecomment-6312382326
Closing this one. The benchmark in the description settles it: turning `open_file_cache` off makes every disk-cache HIT pay one extra `openat` + `fstat` + `close`, about 6 µs of worker CPU per request (roughly +15% CPU for a 1KB cached body, +8% for 64KB). Plain proxying and memory-strategy cache hits are unaffected, but disk-cache hits are exactly the path people pick proxy-cache for, and paying that on every hit to make PURGE take effect immediately isn't a good trade. So the current behaviour stays: with the disk strategy, a PURGE returns 200 and deletes the file, but a worker that already has the file's descriptor cached keeps serving it as a `HIT` until `open_file_cache_valid` (60s by default) runs out. If you need the purged content gone sooner than that, plan for up to a minute of stale hits after a PURGE. -- 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]
