lmcache coordinator#
The lmcache coordinator command launches the LMCache MP coordinator, a
standalone HTTP service that tracks the MP server instances in a deployment. MP
servers register with it and send periodic heartbeats; the coordinator evicts
any instance whose heartbeat lapses past --instance-timeout.
It is the preferred form of python -m lmcache.v1.mp_coordinator, which still
works and accepts the same flags. The process runs in the foreground; stop it
with Ctrl-C.
lmcache coordinator [options]
Quick start#
lmcache coordinator \
--host 0.0.0.0 --port 9300 \
--instance-timeout 30 \
--health-check-interval 10
Options#
Flag |
Description |
|---|---|
|
Bind address for the coordinator’s HTTP server (default: |
|
HTTP port (default: |
|
Seconds without a heartbeat after which an instance is evicted
(default: |
|
Seconds between health-check sweeps; |
|
Seconds between L2 eviction sweeps; |
|
Fraction of tracked keys (by count) to evict per cycle, |
|
Eviction fires when usage reaches this fraction of the quota, |
|
Tokens per KV chunk: the CacheBlend match unit and the unit used to
resolve pin |
|
Token hash algorithm for pin key resolution; must equal the MP servers’
|
|
Index stored chunk content so |
|
Positions between CacheBlend match probes; |
|
Checkpoint the coordinator’s directory, usage view and stream cursors to this file, so a restart resumes instead of starting cold. Unset disables checkpointing. |
|
Seconds between checkpoint writes; |
|
Store operator-set state – L2 pins and per- |
|
Seconds the HTTP server keeps idle connections open before closing
them. Must be greater than the MP servers’ heartbeat interval
(default |
|
Disable OpenTelemetry metrics. Metrics are enabled by default. |
|
Push metrics to the specified OTLP gRPC endpoint. When unset, Prometheus
pull mode exposes |
Configuration#
Every flag is optional; an unset flag keeps the built-in default listed above.
Prometheus pull mode reuses the coordinator’s existing HTTP server; it does not
start a second server or reserve a separate Prometheus port. Metrics-disabled
and OTLP push modes both return HTTP 404 from the local /metrics route.
See Multi-Server Coordination for the active eviction loop.
The coordinator drives fleet-wide L2 eviction by calling each MP
server’s DELETE /l2 endpoint, documented at
Cache Objects And Prefetch.
See Multi-Server Coordination for the coordinator’s architecture, registration protocol, and HTTP API.