dragonfly cache folder is empty after successful preheat

[dusagu]

problem

The folder is empty after preheat. only available when pull the image.

Environment

OS: debian 12
Version: latest
docker compose deploy

89b9e6667f48   dragonflyoss/dfdaemon:latest   "/opt/dragonfly/bin/…"   36 minutes ago   Up 36 minutes (healthy)   0.0.0.0:65000-65002->65000-65002/tcp, 0.0.0.0:65006->65006/tcp, 0.0.0.0:65008->65008/tcp   peer

config

/etc/dragonfly/seed-peer.yml

# daemon alive time, when sets 0s, daemon will not auto exit
# it is useful for longtime running
aliveTime: 0s

# daemon gc task running interval
gcInterval: 1h

# WorkHome is working directory.
# In linux, default value is /usr/local/dragonfly.
# In macos(just for testing), default value is /Users/$USER/.dragonfly.
workHome: ''

# logDir is the log directory.
# In linux, default value is /var/log/dragonfly.
# In macos(just for testing), default value is /Users/$USER/.dragonfly/logs.
logDir: ''

# cacheDir is dynconfig cache directory.
# In linux, default value is /var/cache/dragonfly.
# In macos(just for testing), default value is /Users/$USER/.dragonfly/cache.
cacheDir: ''

# pluginDir is the plugin directory.
# In linux, default value is /usr/local/dragonfly/plugins.
# In macos(just for testing), default value is /Users/$USER/.dragonfly/plugins.
pluginDir: ''

# dataDir is the download data directory.
# In linux, default value is /var/lib/dragonfly.
# In macos(just for testing), default value is /Users/$USER/.dragonfly/data.
dataDir: ''

# when daemon exit, keep peer task data or not
# it is usefully when upgrade daemon service, all local cache will be saved
# default is false
keepStorage: true

# console shows log on console
console: false

# whether to enable debug level logger and enable pprof
verbose: false

# listen port for pprof, only valid when the verbose option is true
# default is -1. If it is 0, pprof will use a random port.
pprof-port: -1

# jaeger endpoint url, like: http://jaeger.dragonfly.svc:14268/api/traces
jaeger: ""

# all addresses of all schedulers
# the schedulers of all daemons should be same in one region or zone.
# daemon will send tasks to a fixed scheduler by hashing the task url and meta data
# caution: only tcp is supported
scheduler:
  manager:
    # get scheduler list dynamically from manager
    enable: true
    # manager service addresses
    netAddrs:
      - type: tcp
        addr: 100.**.**.**:65003
        # scheduler list refresh interval
    refreshInterval: 10m
    seedPeer:
      # Dfdaemon enabled seed peer mode.
      enable: true
      # Seed peer type includes super, strong and weak.
      type: weak
      # Seed peer cluster id.
      clusterID: 1
      keepAlive:
        # Keep alive internal.
        internal: 30s
  # schedule timeout
  scheduleTimeout: 30s
  # when true, only scheduler says back source, daemon can back source
  disableAutoBackSource: false

# Current host info used for scheduler.
host:
  # # Access ip for other peers,
  # # when local ip is different with access ip, advertiseIP should be set.
  advertiseIP: 82.**.**.**
  # Geographical location, separated by "|" characters.
  location: ''
  # IDC deployed by daemon.
  idc: ''
  # Daemon hostname.
  # hostname: ""

# Download service option.
download:
  # Calculate digest when transfer files, set false to save memory.
  calculateDigest: true
  # Total download limit per second.
  totalRateLimit: 2048Mi
  # Per peer task download limit per second.
  perPeerRateLimit: 1024Mi
  # Download piece timeout.
  pieceDownloadTimeout: 30s
  # When request data with range header, prefetch data not in range.
  prefetch: true
  # Golang transport option.
  transportOption:
    # Ddial timeout.
    dialTimeout: 30s
    # Keep alive.
    keepAlive: 30s
    # Same with http.Transport.MaxIdleConns.
    maxIdleConns: 100
    # Same with http.Transport.IdleConnTimeout.
    idleConnTimeout: 90s
    # Same with http.Transport.ResponseHeaderTimeout.
    responseHeaderTimeout: 30s
    # Same with http.Transport.TLSHandshakeTimeout.
    tlsHandshakeTimeout: 30s
    # Same with http.Transport.ExpectContinueTimeout.
    expectContinueTimeout: 30s
  # Concurrent option for back source, default: empty
  # if you want to enable concurrent option, thresholdSize and goroutineCount is enough, keep other options empty is okay.
  concurrent:
    # thresholdSize indicates the threshold to download pieces concurrently.
    thresholdSize: 10M
    # thresholdSpeed indicates the threshold download speed to download pieces concurrently.
    thresholdSpeed: 2M
    # goroutineCount indicates the concurrent goroutine count for every task.
    goroutineCount: 4
    # initBackoff second for every piece failed, default: 0.5.
    initBackoff: 0.5
    # maxBackoff second for every piece failed, default: 3.
    maxBackoff: 3
    # maxAttempts for every piece failed,default: 3.
    maxAttempts: 3
  # Download grpc option.
  downloadGRPC:
    # Security option.
    security:
      insecure: true
      cacert: ''
      cert: ''
      key: ''
      tlsVerify: true
      tlsConfig: null
    # Download service listen address
    # current, only support unix domain socket.
    unixListen:
      # In linux, default value is /var/run/dfdaemon.sock.
      # In macos(just for testing), default value is /tmp/dfdaemon.sock.
      socket: ''
  # Peer grpc option.
  # Peer grpc service send pieces info to other peers.
  peerGRPC:
    security:
      insecure: true
      cacert: ''
      cert: ''
      key: ''
      tlsVerify: true
    tcpListen:
      # # Listen address.
      # listen: 0.0.0.0
      # Listen port, daemon will try to listen,
      # when this port is not available, daemon will try next port.
      port: 65006
      # If want to limit upper port, please use blow format.
#     port:
#       start: 65000
#       end: 65009

# Upload service option.
upload:
  # Upload limit per second.
  rateLimit: 2048Mi
  security:
    insecure: true
    cacert: ''
    cert: ''
    key: ''
    tlsVerify: false
  tcpListen:
    # # Listen address.
    # listen: 0.0.0.0
    # Listen port, daemon will try to listen,
    # when this port is not available, daemon will try next port.
    port: 65008
    # If want to limit upper port, please use blow format.
#   port:
#     start: 65020
#     end: 65029

# Object storage service.
objectStorage:
  # Enable object storage service.
  enable: false
  # Filter is used to generate a unique Task ID by
  # filtering unnecessary query params in the URL,
  # it is separated by & character.
  # When filter: "Expires&Signature&ns", for example:
  #  http://localhost/xyz?Expires=111&Signature=222&ns=docker.io and http://localhost/xyz?Expires=333&Signature=999&ns=docker.io
  # is same task.
  filter: 'Expires&Signature&ns'
  # maxReplicas is the maximum number of replicas of an object cache in seed peers.
  maxReplicas: 3
  # Object storage service security option.
  security:
    insecure: true
    tlsVerify: true
  tcpListen:
    # # Listen address.
    # listen: 0.0.0.0
    # Listen port.
    port: 65004

# Peer task storage option.
storage:
  # Task data expire time,
  # when there is no access to a task data, this task will be gc.
  taskExpireTime: 2160h
  # Storage strategy when process task data.
  # io.d7y.storage.v2.simple : download file to data directory first, then copy to output path, this is default action
  #                           the download file in date directory will be the peer data for uploading to other peers.
  # io.d7y.storage.v2.advance: download file directly to output path with postfix, hard link to final output,
  #                            avoid copy to output path, fast than simple strategy, but:
  #                            the output file with postfix will be the peer data for uploading to other peers
  #                            when user delete or change this file, this peer data will be corrupted.
  # default is io.d7y.storage.v2.simple.
  strategy: io.d7y.storage.v2.simple
  # Disk quota gc threshold, when the quota of all tasks exceeds the gc threshold, the oldest tasks will be reclaimed.
  diskGCThreshold: 100Gi
  # Disk used percent gc threshold, when the disk used percent exceeds, the oldest tasks will be reclaimed.
  # eg, diskGCThresholdPercent=80, when the disk usage is above 80%, start to gc the oldest tasks.
  diskGCThresholdPercent: 80
  # Set to ture for reusing underlying storage for same task id.
  multiplex: true

# Health service option.
health:
  security:
    insecure: true
    cacert: ''
    cert: ''
    key: ''
    tlsVerify: false
  tcpListen:
    # # Listen address.
    # listen: 0.0.0.0
    # Listen port, daemon will try to listen,
    # when this port is not available, daemon will try next port.
    port: 40902
    # If want to limit upper port, please use blow format.
#   port:
#     start: 40901
#     end: 40901

# Proxy service detail option.
proxy:
  # Filter for hash url.
  # when defaultFilter: "Expires&Signature&ns", for example:
  #  http://localhost/xyz?Expires=111&Signature=222&ns=docker.io and http://localhost/xyz?Expires=333&Signature=999&ns=docker.io
  # is same task, it is also possible to override the default filter by adding
  # the X-Dragonfly-Filter header through the proxy.
  defaultFilter: 'Expires&Signature&ns'
  # Tag the task.
  # when the value of the default tag is different,
  # the same download url can be divided into different tasks according to the tag,
  # it is also possible to override the default tag by adding
  # the X-Dragonfly-Tag header through the proxy.
  defaultTag: ''
  security:
    insecure: true
    cacert: ''
    cert: ''
    key: ''
    tlsVerify: false
  tcpListen:
    # namespace stands the linux net namespace, like /proc/1/ns/net.
    # It's useful for running daemon in pod with ip allocated and listening the special port in host net namespace.
    # Linux only.
    namespace: ''
    # # Listen address.
    # listen: 0.0.0.0
    # Listen port, daemon will try to listen,
    # when this port is not available, daemon will try next port.
    port: 65001
    # If want to limit upper port, please use blow format.
  #   port:
  #     start: 65020
  #     end: 65029
  registryMirror:
    # When enable, using header "X-Dragonfly-Registry" for remote instead of url.
    dynamic: true
    # URL for the registry mirror.
    url: https://registry.dcim.co
    # Whether to ignore https certificate errors.
    insecure: true
    # Optional certificates if the remote server uses self-signed certificates.
    certs: []
    # Whether to request the remote registry directly.
    direct: false
    # Whether to use proxies to decide if dragonfly should be used.
    useProxies: false

  proxies:
    # Proxy all http image layer download requests with dfget.
    - regx: blobs/sha256.*
    # Proxy all http image layer download requests with dfget.
    - regx: file-server.*
    # Change http requests to some-registry to https and proxy them with dfget.
    - regx: some-registry/
      useHTTPS: true
    # Proxy requests directly, without dfget.
    - regx: no-proxy-reg
      direct: true
    # Proxy requests with redirect.
    - regx: some-registry
      redirect: another-registry
    # The same with url rewrite like apache ProxyPass directive.
    - regx: ^http://some-registry/(.*)
      redirect: http://another-registry/$1

  hijackHTTPS:
    # key pair used to hijack https requests
    cert: ""
    key: ""
    hosts:
      - regx: mirror.aliyuncs.com:443 # regexp to match request hosts
        # whether to ignore https certificate errors
        insecure: true
        # optional certificates if the host uses self-signed certificates
        certs: []
  # max tasks to download same time, 0 is no limit
  maxConcurrency: 0
  whiteList:
    # the host of the whitelist
    - host: ""
      # match whitelist hosts
      regx: ".*"
      # port that need to be added to the whitelist
      ports:



security:
  # autoIssueCert indicates to issue client certificates for all grpc call.
  # If AutoIssueCert is false, any other option in Security will be ignored.
  autoIssueCert: false
  # caCert is the root CA certificate for all grpc tls handshake, it can be path or PEM format string.
  caCert: ''
  # tlsVerify indicates to verify certificates.
  tlsVerify: false
  # tlsPolicy controls the grpc shandshake behaviors:
  #   force: both ClientHandshake and ServerHandshake are only support tls
  #   prefer: ServerHandshake supports tls and insecure (non-tls), ClientHandshake will only support tls
  #   default: ServerHandshake supports tls and insecure (non-tls), ClientHandshake will only support insecure (non-tls)
  # Notice: If the drgaonfly service has been deployed, a two-step upgrade is required.
  # The first step is to set tlsPolicy to default, and then upgrade the dragonfly services.
  # The second step is to set tlsPolicy to prefer, and then completely upgrade the dragonfly services.
  tlsPolicy: 'prefer'
  certSpec:
    # validityPeriod is the validity period  of certificate.
    validityPeriod: 4320h
# Prometheus metrics address.
# metrics: ':8000'

network:
  # Enable ipv6.
  enableIPv6: false

scheduler.yml

# server scheduler instance configuration
server:
  # # Advertise ip.
  advertiseIP: 100.**.**.**
  # # Listen ip.
  # listenIP: 0.0.0.0
  # Port is the ip and port scheduler server listens on.
  port: 8002
  # # Server host.
  # host: localhost
  # WorkHome is working directory.
  # In linux, default value is /usr/local/dragonfly.
  # In macos(just for testing), default value is /Users/$USER/.dragonfly.
  workHome: ''
  # logDir is the log directory.
  # In linux, default value is /var/log/dragonfly.
  # In macos(just for testing), default value is /Users/$USER/.dragonfly/logs.
  logDir: ''
  # cacheDir is dynconfig cache directory.
  # In linux, default value is /var/cache/dragonfly.
  # In macos(just for testing), default value is /Users/$USER/.dragonfly/cache.
  cacheDir: ''
  # pluginDir is the plugin directory.
  # In linux, default value is /usr/local/dragonfly/plugins.
  # In macos(just for testing), default value is /Users/$USER/.dragonfly/plugins.
  pluginDir: ''
  # dataDir is the directory.
  # In linux, default value is /var/lib/dragonfly.
  # In macos(just for testing), default value is /Users/$USER/.dragonfly/data.
  dataDir: ''

# scheduler policy configuration
scheduler:
  # Algorithm configuration to use different scheduling algorithms,
  # default configuration supports "default" and "ml"
  # "default" is the rule-based scheduling algorithm,
  # "ml" is the machine learning scheduling algorithm
  # It also supports user plugin extension, the algorithm value is "plugin",
  # and the compiled `d7y-scheduler-plugin-evaluator.so` file is added to
  # the dragonfly working directory plugins.
  algorithm: default
  # backSourceCount is the number of backsource clients
  # when the seed peer is unavailable.
  backSourceCount: 3
  # Retry scheduling back-to-source limit times.
  retryBackSourceLimit: 5
  # Retry scheduling limit times.
  retryLimit: 10
  # Retry scheduling interval.
  retryInterval: 50ms
  # GC metadata configuration.
  gc:
    # pieceDownloadTimeout is the timeout of downloading piece.
    pieceDownloadTimeout: 30m
    # peerGCInterval is the interval of peer gc.
    peerGCInterval: 10s
    # peerTTL is the ttl of peer. If the peer has been downloaded by other peers,
    # then PeerTTL will be reset
    peerTTL: 24h
    # taskGCInterval is the interval of task gc. If all the peers have been reclaimed in the task,
    # then the task will also be reclaimed.
    taskGCInterval: 30m
    # hostGCInterval is the interval of host gc.
    hostGCInterval: 6h
    # hostTTL is time to live of host. If host announces message to scheduler,
    # then HostTTl will be reset.
    hostTTL: 1h

# Database info used for server.
database:
  # Redis configuration.
  redis:
    # Redis addresses.
    addrs:
      - "redis:6379"
    # Redis username.
    username: ''
    # Redis password.
    password: dragonfly
    # Redis brokerDB name.
    brokerDB: 1
    # Redis backendDB name.
    backendDB: 2

# Dynamic data configuration.
dynConfig:
  # Dynamic config refresh interval.
  refreshInterval: 1m

# Scheduler host configuration.
host:
  # idc is the idc of scheduler instance.
  idc: ''
  # location is the location of scheduler instance.
  location: ''

# Manager configuration.
manager:
  # addr is manager access address.
  addr: "100.**.**.**.**:65003"
  # schedulerClusterID cluster id to which scheduler instance belongs.
  schedulerClusterID: "1"
  # keepAlive keep alive configuration.
  keepAlive:
    # KeepAlive interval.
    interval: 5s

# Seed peer configuration.
seedPeer:
  # Scheduler enable seed peer as P2P peer,
  # if the value is false, P2P network will not be back-to-source through
  # seed peer but by peer and preheat feature does not work.
  enable: true

# Machinery async job configuration,
# see https://github.com/RichardKnop/machinery.
job:
  # Scheduler enable job service.
  enable: true
  # Number of workers in global queue.
  globalWorkerNum: 1
  # Number of workers in scheduler queue.
  schedulerWorkerNum: 1
  # Number of workers in local queue.
  localWorkerNum: 5

# Store task download information.
storage:
  # maxSize sets the maximum size in megabytes of storage file.
  maxSize: 100
  # maxBackups sets the maximum number of storage files to retain.
  maxBackups: 10
  # bufferSize sets the size of buffer container,
  # if the buffer is full, write all the records in the buffer to the file.
  bufferSize: 100

# Enable prometheus metrics.
metrics:
  # Scheduler enable metrics service.
  enable: false
  # Metrics service address.
  addr: ':8000'
  # Enable host metrics.
  enableHost: false

security:
  # autoIssueCert indicates to issue client certificates for all grpc call.
  # If AutoIssueCert is false, any other option in Security will be ignored.
  autoIssueCert: false
  # caCert is the root CA certificate for all grpc tls handshake, it can be path or PEM format string.
  caCert: ''
  # tlsVerify indicates to verify certificates.
  tlsVerify: false
  # tlsPolicy controls the grpc shandshake behaviors:
  #   force: both ClientHandshake and ServerHandshake are only support tls
  #   prefer: ServerHandshake supports tls and insecure (non-tls), ClientHandshake will only support tls
  #   default: ServerHandshake supports tls and insecure (non-tls), ClientHandshake will only support insecure (non-tls)
  # Notice: If the drgaonfly service has been deployed, a two-step upgrade is required.
  # The first step is to set tlsPolicy to default, and then upgrade the dragonfly services.
  # The second step is to set tlsPolicy to prefer, and then completely upgrade the dragonfly services.
  tlsPolicy: 'prefer'
  certSpec:
    # validityPeriod is the validity period  of certificate.
    validityPeriod: 4320h

network:
  # Enable ipv6.
  enableIPv6: false

# console shows log on console
console: false

# whether to enable debug level logger and enable pprof
verbose: true

# listen port for pprof, only valid when the verbose option is true
# default is -1. If it is 0, pprof will use a random port.
pprof-port: -1

# jaeger endpoint url, like: http://jaeger.dragonfly.svc:14268/api/traces
jaeger: ""

docker-compose.yml

services:
  dfdaemon:
    image: dragonflyoss/dfdaemon:latest
    container_name: peer
    restart: always
    healthcheck:
      test: ["CMD-SHELL", "/bin/grpc_health_probe -addr=:65006 || exit 1"]
      interval: 1s
      timeout: 2s
      retries: 30
    volumes:
      - /var/log/peer:/var/log/dragonfly/daemon
      - /mnt/cache/dragonfly:/var/lib/dragonfly
      - /etc/dragonfly/seed-peer.yaml:/etc/dragonfly/dfget.yaml:ro
    ports:
      - 65000:65000
      - 65001:65001
      - 65002:65002
      - 65006:65006
      - 65008:65008
    extra_hosts:
      - "harbor.**.**:100.**.**.**"

[gaius-qi]

@dusagu Please pull and merge the latest version of the main branch. Docker compose has been updated maybe two months ago.

[dusagu]

@gaius-qi
But here it is 5hours ago

[gaius-qi]

@gaius-qi

But here it is 5hours ago

Update the config of the docker compose.