Customer Server Lifecycle - Order to Termination

Customer Server Lifecycle - Order to Termination A lifecycle diagram generated by Archify. 01 / Lifecycle phases 02 / Wake path (free tier only) + Recoverable states 03 / Terminal exit Ordered · order or free claim · Lifecycle phases · entry 01 Ordered order or free claim entry Installing · egg installer in Docker · Lifecycle phases 02 Installing egg installer in Docker Active · players connected · Lifecycle phases · running 03 Active players connected running Idle check · hourly · inactive ≥ 2 days · Lifecycle phases 04 Idle check hourly · inactive ≥ 2 days Asleep · port owned by waker · Lifecycle phases · most servers 05 Asleep port owned by waker most servers In limbo · waker holds the socket · Wake path (free tier only) · ≤ 60 s In limbo waker holds the socket ≤ 60 s install_failed · often a silently empty volume · Recoverable states · reinstallable install_failed often a silently empty volume reinstallable Suspended · abuse or non-payment · Recoverable states · reversible Suspended abuse or non-payment reversible Terminated · volume deleted · Terminal exit · irreversible Terminated volume deleted irreversible installer exits non-zero reinstall player connects port hand-off abuse or billing appeal upheld / paid retention expires Legend start active state waiting decision failure / exit

The free tier lives asleep

  • • activity-checker stops servers idle for 2 days, at most 20 per hourly pass, with 24 h warning
  • • Most of node001's 1487 servers are stopped at any moment
  • • node002 is the exception: paid servers carry no sleep/wake automation and never enter Idle or Asleep
  • • Only a player connecting wakes a server - no schedule, no panel poll

Why installs fail silently

  • • node001's /etc/docker/daemon.json sets init-path to a KSM shim and /sbin/docker-init does not exist
  • • The egg installer dies with 'exec /sbin/docker-init' and the panel still records a server
  • • The result is a server whose volume is empty rather than an obvious error
  • • A customer-driven jar or egg switch can also leave an unbootable world; check activity_logs for server:version.install first

Suspension is reversible; termination is not

  • • AbuseShield suspends only HIGH-confidence findings, never MEDIUM
  • • Archive the evidence before any delete - the flag store is the only record
  • • A stale expires_at in Paymenter once made the cron suspend AND terminate in the same pass, destroying a paid server
  • • Erasure order matters: email_logs.user_id is SET NULL, not CASCADE, so it orphans the address and message bodies