/opt/chainguard/sglang/bin/mooncake_master

mooncake_master: Warning: SetUsageMessage() never called

  Flags from /__w/Mooncake/Mooncake/mooncake-store/src/master.cpp:
    -allocation_strategy (Allocation strategy for segments, random |
      free_ratio_first | cxl | ssd_free_ratio_first | local_first) type: string
      default: "random"
    -allow_evict_soft_pinned_objects (Whether to allow eviction of soft pinned
      objects during eviction) type: bool default: true
    -batch_oplog_retry_timeout_sec (Maximum time to retry transient batch OpLog
      standby errors.) type: uint32 default: 180
    -client_ttl (Seconds a client stays considered alive after the last
      heartbeat. If this TTL elapses without a refresh, the master treats the
      client as disconnected and may unmount its segments) type: int64
      default: 10
    -cluster_id (Cluster ID for the master service, used for kvcache
      persistence in HA mode) type: string default: "mooncake_cluster"
    -config_path (master service config file path) type: string default: ""
    -cxl_path (DAX device path for CXL memory) type: string
      default: "/dev/dax0.0"
    -cxl_size (CXL memory size in bytes) type: uint64 default: 8589934592
    -default_kv_lease_ttl (Default lease time for kv objects. Supports raw
      milliseconds or duration strings with ms, s, m, or h suffixes)
      type: string default: "10000"
    -default_kv_soft_pin_ttl (Default soft pin TTL for kv objects. Supports raw
      milliseconds or duration strings with ms, s, m, or h suffixes)
      type: string default: "1800000"
    -dynamic_replication_admission_qps_threshold (Per-key access frequency
      threshold for dynamic MEMORY replication) type: double
      default: 0.80000000000000004
    -dynamic_replication_heat_window_seconds (Heat observation window for
      dynamic MEMORY replication) type: uint32 default: 10
    -dynamic_replication_max_memory_replicas (Maximum MEMORY replicas allowed
      for one dynamically replicated key) type: uint64 default: 2
    -dynamic_replication_mode (Dynamic MEMORY replica fanout mode: off,
      observe, or enforce) type: string default: "off"
    -enable_cxl (Whether to enable CXL memory support) type: bool
      default: false
    -enable_disk_eviction (Enable disk eviction feature for storage backend
      (default: true)) type: bool default: true
    -enable_ha (Enable high availability using the configured HA backend)
      type: bool default: false
    -enable_http_metadata_server (Enable HTTP metadata server instead of etcd)
      type: bool default: false
    -enable_kv_events (Enable RFC #1527 KV cache event publisher over ZMQ)
      type: bool default: false
    -enable_metadata_cleanup_on_timeout (Enable cleanup of HTTP metadata
      (mooncake/ram/*, mooncake/rpc_meta/*) when client heartbeat times out.
      Works in two modes: (1) co-located (enable_http_metadata_server=true) via
      in-process removal, or (2) separately-deployed metadata server via async
      HTTP DELETE.) type: bool default: false
    -enable_metric_reporting (Enable periodic metric reporting) type: bool
      default: true
    -enable_multi_tenants (Enable strict multi-tenant namespace and quota
      admission) type: bool default: false
    -enable_offload (Enable offload availability) type: bool default: false
    -enable_oplog (Enable HA metadata replication through batch-record OpLog)
      type: bool default: false
    -enable_snapshot (Enable periodic snapshot of master data) type: bool
      default: false
    -enable_snapshot_restore (enable restore from snapshot) type: bool
      default: false
    -etcd_endpoints (Endpoints of ETCD server, separated by semicolon, required
      in HA mode) type: string default: ""
    -eviction_high_watermark_ratio (Ratio of high watermark trigger eviction in
      Memory) type: double default: 0.90000000000000002
    -eviction_ratio (Ratio of objects to evict when Memory space is full)
      type: double default: 0.050000000000000003
    -global_file_segment_size (Size of global NFS/3FS segment in bytes)
      type: int64 default: 9223372036854775807
    -ha_backend_connstring (HA backend connection string. If unset, only
      backend_type=etcd falls back to etcd_endpoints for backward
      compatibility) type: string default: ""
    -ha_backend_type (HA backend type, e.g. etcd | redis | k8s) type: string
      default: "etcd"
    -http_metadata_server_host (Host for HTTP metadata server to bind to)
      type: string default: "0.0.0.0"
    -http_metadata_server_port (Port for HTTP metadata server to listen on)
      type: int32 default: 8080
    -kv_events_additional_salt (Deprecated: not emitted on events; use indexer
      POST /register) type: string default: ""
    -kv_events_backend_id (backend_id for published KV events (cache owner
      identity)) type: string default: ""
    -kv_events_bind_endpoint (ZMQ PUB bind endpoint for KV events, e.g.
      tcp://0.0.0.0:5557) type: string default: ""
    -kv_events_block_size (Deprecated: not emitted on events; use indexer POST
      /register) type: uint32 default: 0
    -kv_events_dp_rank (Deprecated: not emitted on events; use indexer POST
      /register) type: uint32 default: 0
    -kv_events_emit_legacy_compat (Include vLLM/SGLang-compatible
      type/block_hashes fields) type: bool default: true
    -kv_events_emit_object_key (Include Mooncake object_key in published KV
      events) type: bool default: true
    -kv_events_lora_name (Deprecated: not emitted on events (no LoRA context in
      master)) type: string default: ""
    -kv_events_model_name (Deprecated: not emitted on events; use indexer POST
      /register) type: string default: ""
    -kv_events_queue_capacity (Maximum pending KV events; oldest events are
      dropped when the queue is full (0 = unbounded)) type: uint32
      default: 65536
    -kv_events_tenant_id (Deprecated: tenant_id comes from each object on
      events) type: string default: "default"
    -max_kv_soft_pin_ttl (Maximum request-level soft pin TTL for kv objects.
      Supports raw milliseconds or duration strings with ms, s, m, or h
      suffixes) type: string default: "86400000"
    -max_retry_attempts (Maximum number of retry attempts for failed tasks)
      type: uint32 default: 10
    -max_threads (Maximum number of threads to use (deprecated, use
      rpc_thread_num)) type: int32 default: 16
    -max_total_finished_tasks (Maximum number of finished tasks to keep in
      memory) type: uint32 default: 10000
    -max_total_pending_tasks (Maximum number of pending tasks to keep in
      memory) type: uint32 default: 10000
    -max_total_processing_tasks (Maximum number of processing tasks to keep in
      memory) type: uint32 default: 10000
    -memory_allocator (Memory allocator for global segments, cachelib | offset)
      type: string default: "offset"
    -metrics_host (Address for the HTTP metrics/admin server to listen on. Use
      "::" to listen on IPv6 (and IPv4 on dual-stack hosts)) type: string
      default: "0.0.0.0"
    -metrics_port (Port for HTTP metrics server to listen on) type: int32
      default: 9003
    -nof_eviction_high_watermark_ratio (Ratio of high watermark trigger
      eviction in NoF SSD) type: double default: 0.90000000000000002
    -nof_eviction_ratio (Ratio of objects to evict when NoF SSD space is full)
      type: double default: 0.050000000000000003
    -nof_heartbeat_failures_threshold (Consecutive NoF heartbeat failures
      required before unmounting a NoF segment) type: uint32 default: 3
    -nof_heartbeat_interval_sec (How often master probes each mounted NoF
      segment) type: int64 default: 10
    -nof_heartbeat_probe_timeout_ms (Timeout in milliseconds for a single NoF
      heartbeat probe) type: uint32 default: 1000
    -offload_cap_ratio (Per-cycle offload cap as a fraction of
      offloading_queue_limit. Controls how many objects can be queued for
      offload in a single eviction cycle before falling back to force-evict)
      type: double default: 0.5
    -offload_force_evict (Force-evict objects exceeding offload cap without
      disk offload) type: bool default: false
    -offload_on_evict (Defer LOCAL_DISK offload to eviction time instead of
      PutEnd) type: bool default: false
    -offloading_queue_limit (Maximum number of objects allowed in the
      offloading queue per local disk segment. Increase to allow more objects
      to be offloaded to SSD before force-eviction kicks in) type: uint64
      default: 50000
    -oplog_batch_max_entries (Maximum number of committed/reserved entries in
      the open batch-record OpLog waiting batch.) type: uint32 default: 1024
    -oplog_poll_interval_ms (Batch-record standby poll interval.) type: int32
      default: 1000
    -pending_task_timeout_sec (Timeout in seconds for pending tasks (0 = no
      timeout)) type: uint64 default: 300
    -pod_name (Pod name for K8s label-based routing (default: $POD_NAME))
      type: string default: ""
    -pod_namespace (Pod namespace for K8s label-based routing (default:
      $POD_NAMESPACE)) type: string default: ""
    -port (Port for master service to listen on (deprecated, use rpc_port))
      type: int32 default: 50051
    -processing_task_timeout_sec (Timeout in seconds for processing tasks (0 =
      no timeout)) type: uint64 default: 300
    -promotion_admission_threshold (Min CountMinSketch count for a key before
      promotion fires (set 1 to disable second-touch gating)) type: uint32
      default: 2
    -promotion_max_per_heartbeat (Max promotion tasks returned to a single
      client per PromotionObjectHeartbeat call. Each task is a synchronous
      SSD-read + RDMA-write on the client; serializing them avoids blocking
      past the client-liveness window. Default 1 is conservative.) type: uint32
      default: 1
    -promotion_on_hit (Promote LOCAL_DISK-only keys to MEMORY on read access
      (mirror of offload_on_evict)) type: bool default: false
    -promotion_queue_limit (Max in-flight promotion tasks across all shards)
      type: uint32 default: 50000
    -put_start_discard_timeout_sec (Timeout for discarding uncompleted PutStart
      operations) type: uint64 default: 30
    -put_start_release_timeout_sec (Timeout for releasing space allocated in
      uncompleted PutStart operations) type: uint64 default: 600
    -quota_bytes (Quota for storage backend in bytes (0 = use default 90% of
      capacity)) type: uint64 default: 0
    -root_fs_dir (Root directory for storage backend, used in HA mode)
      type: string default: ""
    -rpc_address (Address for RPC server to bind to, required in HA mode)
      type: string default: "0.0.0.0"
    -rpc_conn_timeout_seconds (Connection timeout in seconds (0 = no timeout))
      type: int32 default: 0
    -rpc_enable_tcp_no_delay (Enable TCP_NODELAY for RPC connections)
      type: bool default: true
    -rpc_interface (Network interface name for RPC server address resolution.
      When set, its IPv4 address overrides rpc_address) type: string
      default: ""
    -rpc_port (Port for RPC server to listen on (0 = use port, preferred over
      port)) type: int32 default: 0
    -rpc_thread_num (Number of threads for RPC server (0 = use max_threads,
      preferred over max_threads)) type: int32 default: 0
    -snapshot_backup_dir (Optional local directory for snapshot and restore
      backup. If empty, local backup is disabled) type: string default: ""
    -snapshot_catalog_backend_connstring (Deprecated alias of
      --snapshot_catalog_store_connstring) type: string default: ""
    -snapshot_catalog_backend_type (Deprecated alias of
      --snapshot_catalog_store_type) type: string default: ""
    -snapshot_catalog_store_connstring (Optional connection string for snapshot
      catalog store) type: string default: ""
    -snapshot_catalog_store_type (Snapshot catalog store type: ''/'embedded' or
      'redis' ('payload' is kept as a deprecated alias)) type: string
      default: ""
    -snapshot_child_timeout_seconds (Timeout for snapshot child process in
      seconds) type: uint64 default: 300
    -snapshot_interval_seconds (Interval in second between periodic snapshots
      of master data) type: uint64 default: 600
    -snapshot_object_store_type (Snapshot object store type: 'local' for local
      filesystem, 's3' for S3 storage) type: string default: ""
    -snapshot_payload_backend_type (Deprecated alias of
      --snapshot_object_store_type) type: string default: ""
    -snapshot_payload_store_type (Deprecated alias of
      --snapshot_object_store_type) type: string default: ""
    -snapshot_retention_count (Number of recent snapshots to keep (older
      snapshots will be automatically deleted)) type: uint32 default: 2
    -tenant_quota_connector_type (Tenant quota policy connector type)
      type: string default: "file"
    -tenant_quota_connector_uri (Tenant quota policy connector URI)
      type: string default: ""

  Flags from /__w/Mooncake/Mooncake/mooncake-store/src/utils.cpp:
    -mmap_arena_pool_size (Arena allocator pool size in bytes) type: uint64
      default: 8589934592
    -use_mmap_arena_allocator (Enable the lock-free mmap arena allocator for
      mmap buffers) type: bool default: false



  Flags from /builddir/build/BUILD/gflags-2.1.2/src/gflags.cc:
    -flagfile (load flags from file) type: string default: ""
    -fromenv (set flags from the environment [use 'export FLAGS_flag1=value'])
      type: string default: ""
    -tryfromenv (set flags from the environment if present) type: string
      default: ""
    -undefok (comma-separated list of flag names that it is okay to specify on
      the command line even if the program does not define a flag with that
      name.  IMPORTANT: flags in this list that have arguments MUST use the
      flag=value format) type: string default: ""

  Flags from /builddir/build/BUILD/gflags-2.1.2/src/gflags_completions.cc:
    -tab_completion_columns (Number of columns to use in output for tab
      completion) type: int32 default: 80
    -tab_completion_word (If non-empty, HandleCommandLineCompletions() will
      hijack the process and attempt to do bash-style command line flag
      completion on this value.) type: string default: ""

  Flags from /builddir/build/BUILD/gflags-2.1.2/src/gflags_reporting.cc:
    -help (show help on all flags [tip: all flags can have two dashes])
      type: bool default: false currently: true
    -helpfull (show help on all flags -- same as -help) type: bool
      default: false
    -helpmatch (show help on modules whose name contains the specified substr)
      type: string default: ""
    -helpon (show help on the modules named by this flag value) type: string
      default: ""
    -helppackage (show help on all modules in the main package) type: bool
      default: false
    -helpshort (show help on only the main module for this program) type: bool
      default: false
    -helpxml (produce an xml version of help) type: bool default: false
    -version (show version and build info and exit) type: bool default: false



  Flags from src/logging.cc:
    -alsologtoemail (log messages go to these email addresses in addition to
      logfiles) type: string default: ""
    -alsologtostderr (log messages go to stderr in addition to logfiles)
      type: bool default: false
    -colorlogtostderr (color messages logged to stderr (if supported by
      terminal)) type: bool default: false
    -drop_log_memory (Drop in-memory buffers of log contents. Logs can grow
      very quickly and they are rarely read before they need to be evicted from
      memory. Instead, drop them from memory as soon as they are flushed to
      disk.) type: bool default: true
    -log_backtrace_at (Emit a backtrace when logging at file:linenum.)
      type: string default: ""
    -log_dir (If specified, logfiles are written into this directory instead of
      the default logging directory.) type: string default: ""
    -log_link (Put additional links to the log files in this directory)
      type: string default: ""
    -log_prefix (Prepend the log prefix to the start of each log line)
      type: bool default: true
    -logbuflevel (Buffer log messages logged at this level or lower (-1 means
      don't buffer; 0 means buffer INFO only; ...)) type: int32 default: 0
    -logbufsecs (Buffer log messages for at most this many seconds) type: int32
      default: 30
    -logemaillevel (Email log messages logged at this level or higher (0 means
      email all; 3 means email FATAL only; ...)) type: int32 default: 999
    -logfile_mode (Log file mode/permissions.) type: int32 default: 436
    -logmailer (Mailer used to send logging email) type: string
      default: "/bin/mail"
    -logtostderr (log messages go to stderr instead of logfiles) type: bool
      default: false
    -max_log_size (approx. maximum log file size (in MB). A value of 0 will be
      silently overridden to 1.) type: int32 default: 1800
    -minloglevel (Messages logged at a lower level than this don't actually get
      logged anywhere) type: int32 default: 0
    -stderrthreshold (log messages at or above this level are copied to stderr
      in addition to logfiles.  This flag obsoletes --alsologtostderr.)
      type: int32 default: 2
    -stop_logging_if_full_disk (Stop attempting to log to disk if the disk is
      full.) type: bool default: false

  Flags from src/utilities.cc:
    -symbolize_stacktrace (Symbolize the stack trace in the tombstone)
      type: bool default: true

  Flags from src/vlog_is_on.cc:
    -v (Show all VLOG(m) messages for m <= this. Overridable by --vmodule.)
      type: int32 default: 0
    -vmodule (per-module verbose level. Argument is a comma-separated list of
      <module name>=<log level>. <module name> is a glob pattern, matched
      against the filename base (that is, name ignoring .cc/.h./-inl.h). <log
      level> overrides any value given by --v.) type: string default: ""