LMCache · Issues· 757 open

MP mode (LMCacheMPConnector) ignores save_decode_cache and stores decode KV, causing unbounded L1/L2 writes and premature eviction

[good-first-issue] v1: convert cache-controller logging to %-format

[good-first-issue] v1: convert API and server entrypoint logging to %-format

[good-first-issue] tests: convert disagg cache-engine logging to %-format

[good-first-issue] logging: convert top-level LMCache utility logging to %-format

[good-first-issue] benchmarks: convert rag logging to %-format

[good-first-issue] benchmarks: convert multi_round_qa logging to %-format

[Bug] MP server: lookup reports full hit but restored KV context is wrong for long prefixes (breaks between 114 and 128 chunks)

[good-first-issue] buildkite: convert async_request logging to %-format

[good-first-issue] v1: convert config and cache-engine logging to %-format

[good-first-issue] connectors: convert HF3FS adapter logging to %-format

[Feature][MP] Negotiate chunk_size from vLLM KV geometry at first registration

[good-first-issue] examples: convert disagg_prefill logging to %-format

[good-first-issue] storage: convert PD backend logging to %-format

[good-first-issue] compute: convert blend and model logging to %-format