The firstAvailable policy uses QPS limits per server. These do not work as intended if a cache is attached to a pool: we first pick a server and THEN consult the cache. This means the server selection policy sees the uncached QPS, not the cached QPS.
This not alone breaks the QPS but also causes additional work.
Environment
dnsdist from master
Steps to reproduce
Configure backends with QPS, configure a cache, send ton of cacheable traffic.
The text was updated successfully, but these errors were encountered:
Short description
The firstAvailable policy uses QPS limits per server. These do not work as intended if a cache is attached to a pool: we first pick a server and THEN consult the cache. This means the server selection policy sees the uncached QPS, not the cached QPS.
This not alone breaks the QPS but also causes additional work.
Environment
dnsdist from master
Steps to reproduce
Configure backends with QPS, configure a cache, send ton of cacheable traffic.
The text was updated successfully, but these errors were encountered: