Documentation
¶
Index ¶
Constants ¶
View Source
const ( DIR_NAME = ".cedana" FILE_NAME = "config" FILE_TYPE = "json" DIR_PERM = 0o755 FILE_PERM = 0o644 ENV_PREFIX = "CEDANA" // NOTE: `localhost` server inside kubernetes may or may not work // based on firewall and network configuration, it would only work // on local system, hence for serving on TCP default is 0.0.0.0 DEFAULT_TCP_ADDR = "0.0.0.0:8080" DEFAULT_SOCK_ADDR = "/run/cedana.sock" DEFAULT_SOCK_PERMS = 0o666 DEFAULT_PROTOCOL = "unix" DEFAULT_LOG_LEVEL = "info" DEFAULT_LOG_LEVEL_NO_SERVER = "warn" DEFAULT_CHECKPOINT_COMPRESSION = "none" DEFAULT_CHECKPOINT_DIR = "/tmp" DEFAULT_CHECKPOINT_STREAMS = 0 DEFAULT_DB_REMOTE = false DEFAULT_DB_PATH = "/tmp/cedana.db" DEFAULT_PROFILING_ENABLED = true DEFAULT_PROFILING_PRECISION = "auto" DEFAULT_CONNECTION_URL = "https://sandbox.cedana.ai" DEFAULT_CONNECTION_AUTH_TOKEN = "" DEFAULT_METRICS = false DEFAULT_CLIENT_WAIT_FOR_READY = false DEFAULT_GPU_POOL_SIZE = 0 DEFAULT_GPU_LOG_DIR = "/tmp" DEFAULT_GPU_SOCK_DIR = "/tmp" DEFAULT_GPU_FREEZE_TYPE = "IPC" DEFAULT_GPU_SHM_SIZE = 8 * utils.GIBIBYTE DEFAULT_GPU_DEBUG = false DEFAULT_CRIU_LEAVE_RUNNING = false DEFAULT_CRIU_MANAGE_CGROUPS = "default" DEFAULT_PLUGINS_LIB_DIR = "/usr/local/lib" DEFAULT_PLUGINS_BIN_DIR = "/usr/local/bin" DEFAULT_PLUGINS_BUILDS = "release" )
Variables ¶
This section is empty.
Functions ¶
Types ¶
type AWS ¶ added in v0.9.257
type AWS struct {
// AccessKeyID is the AWS access key ID
AccessKeyID string `json:"access_key_id" key:"access_key_id" yaml:"access_key_id" mapstructure:"access_key_id" env_aliases:"AWS_ACCESS_KEY_ID"`
// SecretAccessKey is the AWS secret access key
SecretAccessKey string `` /* 142-byte string literal not displayed */
// Region is the AWS region to use (uses default region if not set)
Region string `json:"region" key:"region" yaml:"region" mapstructure:"region" env_aliases:"AWS_REGION"`
}
type CRIU ¶
type CRIU struct {
// BinaryPath is a custom path to the CRIU binary
BinaryPath string `json:"binary_path" key:"binary_path" yaml:"binary_path" mapstructure:"binary_path"`
// LeaveRunning sets whether to leave the process running after checkpoint
LeaveRunning bool `json:"leave_running" key:"leave_running" yaml:"leave_running" mapstructure:"leave_running"`
// ManageCgroups sets the default cgroup C/R mode for CRIU (default, cg_none, props, soft, full, strict, ignore)
ManageCgroups string `json:"manage_cgroups" key:"manage_cgroups" yaml:"manage_cgroups" mapstructure:"manage_cgroups"`
}
type Checkpoint ¶
type Checkpoint struct {
// Dir is the default directory to store checkpoints
// - "cedana://<path>" for Cedana-managed global storage (recommended)
// - "s3://<path>" for your S3 storage
// - "<path>" for node-local storage
Dir string `json:"dir" key:"dir" yaml:"dir" mapstructure:"dir"`
// Compression is the default compression algorithm to use for checkpoints
Compression string `json:"compression" key:"compression" yaml:"compression" mapstructure:"compression"`
// Streams specifies the number of parallel streams to use when checkpointing.
Streams int32 `json:"streams" key:"streams" yaml:"streams" mapstructure:"streams"`
}
type Client ¶
type Client struct {
// Wait for ready ensures client requests block if the daemon is not up yet
WaitForReady bool `` /* 130-byte string literal not displayed */
}
type Config ¶
type Config struct {
// Address to use for incoming/outgoing connections
Address string `json:"address" key:"address" yaml:"address" mapstructure:"address"`
// Protocol to use for incoming/outgoing connections (TCP, UNIX, VSOCK)
Protocol string `json:"protocol" key:"protocol" yaml:"protocol" mapstructure:"protocol"`
// LogLevel is the default log level used by the server
LogLevel string `json:"log_level" key:"log_level" yaml:"log_level" mapstructure:"log_level"`
// LogLevelNoServer is the log level used when direct --no-server run/restore is used. This is separate from LogLevel so as to avoid cluttering the process output.
LogLevelNoServer string `json:"log_level_no_server" key:"log_level_no_server" yaml:"log_level_no_server" mapstructure:"log_level_no_server"`
// Metrics is whether to enable metrics collection and observability
Metrics bool `json:"metrics" key:"metrics" yaml:"metrics" mapstructure:"metrics"`
// Connection settings
Connection Connection `json:"connection" key:"connection" yaml:"connection" mapstructure:"connection"`
// Checkpoint and storage settings
Checkpoint Checkpoint `json:"checkpoint" key:"checkpoint" yaml:"checkpoint" mapstructure:"checkpoint"`
// Database details
DB DB `json:"db" key:"db" yaml:"db" mapstructure:"db"`
// Profiling settings
Profiling Profiling `json:"profiling" key:"profiling" yaml:"profiling" mapstructure:"profiling"`
// Client settings
Client Client `json:"client" key:"client" yaml:"client" mapstructure:"client"`
// CRIU settings and defaults
CRIU CRIU `json:"criu" key:"criu" yaml:"criu" mapstructure:"criu"`
// GPU is settings for the GPU plugin
GPU GPU `json:"gpu" key:"gpu" yaml:"gpu" mapstructure:"gpu"`
// Plugin settings
Plugins Plugins `json:"plugins" key:"plugins" yaml:"plugins" mapstructure:"plugins"`
// AWS settings
AWS AWS `json:"aws" key:"aws" yaml:"aws" mapstructure:"aws"`
}
Cedana configuration. Each of the below fields can also be set through an environment variable with the same name, prefixed, and in uppercase. E.g. `Checkpoint.Dir` can be set with `CEDANA_CHECKPOINT_DIR`. The `env_aliases` tag below specifies alternative (alias) environment variable names (comma-separated).
var Global Config = Config{ Protocol: DEFAULT_PROTOCOL, LogLevel: DEFAULT_LOG_LEVEL, LogLevelNoServer: DEFAULT_LOG_LEVEL_NO_SERVER, Metrics: DEFAULT_METRICS, Checkpoint: Checkpoint{ Dir: DEFAULT_CHECKPOINT_DIR, Compression: DEFAULT_CHECKPOINT_COMPRESSION, Streams: DEFAULT_CHECKPOINT_STREAMS, }, DB: DB{ Remote: DEFAULT_DB_REMOTE, Path: DEFAULT_DB_PATH, }, Profiling: Profiling{ Enabled: DEFAULT_PROFILING_ENABLED, Precision: DEFAULT_PROFILING_PRECISION, }, Connection: Connection{ URL: DEFAULT_CONNECTION_URL, AuthToken: DEFAULT_CONNECTION_AUTH_TOKEN, }, Client: Client{ WaitForReady: DEFAULT_CLIENT_WAIT_FOR_READY, }, GPU: GPU{ PoolSize: DEFAULT_GPU_POOL_SIZE, LogDir: DEFAULT_GPU_LOG_DIR, SockDir: DEFAULT_GPU_SOCK_DIR, FreezeType: DEFAULT_GPU_FREEZE_TYPE, ShmSize: DEFAULT_GPU_SHM_SIZE, Debug: DEFAULT_GPU_DEBUG, }, CRIU: CRIU{ LeaveRunning: DEFAULT_CRIU_LEAVE_RUNNING, ManageCgroups: DEFAULT_CRIU_MANAGE_CGROUPS, }, Plugins: Plugins{ LibDir: DEFAULT_PLUGINS_LIB_DIR, BinDir: DEFAULT_PLUGINS_BIN_DIR, Builds: DEFAULT_PLUGINS_BUILDS, }, }
The default global config. This will get overwritten by the config file or env vars during startup, if they exist.
type Connection ¶
type Connection struct {
// URL is your unique Cedana endpoint URL
URL string `json:"url" key:"url" yaml:"url" mapstructure:"url" env_aliases:"CEDANA_URL"`
// AuthToken is your authentication token for the Cedana endpoint
AuthToken string `json:"auth_token" key:"auth_token" yaml:"auth_token" mapstructure:"auth_token" env_aliases:"CEDANA_AUTH_TOKEN"`
}
type DB ¶
type DB struct {
// Remote sets whether to use a remote database
Remote bool `json:"remote" key:"remote" yaml:"remote" mapstructure:"remote" env_aliases:"CEDANA_REMOTE"`
// Path is the local path to the database file. E.g. /tmp/cedana.db
Path string `json:"path" key:"path" yaml:"path" mapstructure:"path"`
}
type GPU ¶
type GPU struct {
// Number of warm GPU controllers to keep in pool
PoolSize int `json:"pool_size" key:"pool_size" yaml:"pool_size" mapstructure:"pool_size"`
// LogDir is the directory to write GPU logs to
LogDir string `json:"log_dir" key:"log_dir" yaml:"log_dir" mapstructure:"log_dir"`
// SockDir is the directory to use for the GPU sockets
SockDir string `json:"sock_dir" key:"sock_dir" yaml:"sock_dir" mapstructure:"sock_dir"`
// Track metrics associated with observability
Observability bool `json:"observability" key:"observability" yaml:"observability" mapstructure:"observability"`
// FreezeType is the type of freeze to use for GPU processes (IPC, NCCL)
FreezeType string `json:"freeze_type" key:"freeze_type" yaml:"freeze_type" mapstructure:"freeze_type"`
// ShmSize is the size in bytes of the shared memory segment to use for GPU processes
ShmSize int64 `json:"shm_size" key:"shm_size" yaml:"shm_size" mapstructure:"shm_size"`
// LdLibPath holds any additional directories to search for GPU libraries
LdLibPath string `json:"ld_lib_path" key:"ld_lib_path" yaml:"ld_lib_path" mapstructure:"ld_lib_path"`
// Debug enables debugging capabilities for the GPU plugin. Daemon will try to attach to existing running GPU controllers
Debug bool `json:"debug" key:"debug" yaml:"debug" mapstructure:"debug"`
}
type Plugins ¶
type Plugins struct {
// BinDir is the directory where plugin binaries are stored
BinDir string `json:"bin_dir" key:"bin_dir" yaml:"bin_dir" mapstructure:"bin_dir"`
// LibDir is the directory where plugin libraries are stored
LibDir string `json:"lib_dir" key:"lib_dir" yaml:"lib_dir" mapstructure:"lib_dir" env_aliases:"CEDANA_PLUGINS_LIB_DIR"`
// Builds is the build versions to list/download for plugins (release, alpha)
Builds string `json:"builds" key:"builds" yaml:"builds" mapstructure:"builds"`
}
type Profiling ¶
type Profiling struct {
// Enabled sets whether to enable and show profiling information
Enabled bool `json:"enabled" key:"enabled" yaml:"enabled" mapstructure:"enabled"`
// Precision sets the time precision when printing profiling information (auto, ns, us, ms, s)
Precision string `json:"precision" key:"precision" yaml:"precision" mapstructure:"precision"`
}
Click to show internal directories.
Click to hide internal directories.