Agent Recovery

If the mesos-agent process on a host exits (perhaps due to a Mesos bug or because the operator kills the process while upgrading Mesos), any executors/tasks that were being managed by the mesos-agent process will continue to run.

By default, all the executors/tasks that were being managed by the old mesos-agent process are expected to gracefully exit on their own, and will be shut down after the agent restarted if they did not.

However, if a framework enabled checkpointing when it registered with the master, any executors belonging to that framework can reconnect to the new mesos-agent process and continue running uninterrupted. Hence, enabling framework checkpointing allows tasks to tolerate Mesos agent upgrades and unexpected mesos-agent crashes without experiencing any downtime.

Agent recovery works by having the agent checkpoint information about its own state and about the tasks and executors it is managing to local disk, for example the SlaveInfo, FrameworkInfo and ExecutorInfo messages or the unacknowledged status updates of running tasks.

When the agent restarts, it will verify that its current configuration, set from the environment variables and command-line flags, is compatible with the checkpointed information and will refuse to restart if not.

A special case occurs when the agent detects that its host system was rebooted since the last run of the agent: The agent will try to recover its previous ID as usual, but if that fails it will actually erase the information of the previous run and will register with the master as a new agent.

Note that executors and tasks that exited between agent shutdown and restart are not automatically restarted during agent recovery.

Framework Configuration

A framework can control whether its executors will be recovered by setting the checkpoint flag in its FrameworkInfo when registering with the master. Enabling this feature results in increased I/O overhead at each agent that runs tasks launched by the framework. By default, frameworks do not checkpoint their state.

Agent Configuration

Four configuration flags control the recovery behavior of a Mesos agent:

  • strict: Whether to do agent recovery in strict mode [Default: true].

    • If strict=true, all recovery errors are considered fatal.
    • If strict=false, any errors (e.g., corruption in checkpointed data) during recovery are ignored and as much state as possible is recovered.

  • reconfiguration_policy: Which kind of configuration changes are accepted when trying to recover [Default: equal].

    • If reconfiguration_policy=equal, no configuration changes are accepted.
    • If reconfiguration_policy=additive, the agent will allow the new configuration to contain additional attributes, increased resourced or an additional fault domain. For a more detailed description, see this.

  • recover: Whether to recover status updates and reconnect with old executors [Default: reconnect]

    • If recover=reconnect, reconnect with any old live executors, provided the executor's framework enabled checkpointing.
    • If recover=cleanup, kill any old live executors and exit. Use this option when doing an incompatible agent or executor upgrade! NOTE: If no checkpointing information exists, no recovery is performed and the agent registers with the master as a new agent.

  • recovery_timeout: Amount of time allotted for the agent to recover [Default: 15 mins].

    • If the agent takes longer than recovery_timeout to recover, any executors that are waiting to reconnect to the agent will self-terminate. NOTE: If none of the frameworks have enabled checkpointing, the executors and tasks running at an agent die when the agent dies and are not recovered.

A restarted agent should reregister with master within a timeout (75 seconds by default: see the --max_agent_ping_timeouts and --agent_ping_timeout configuration flags). If the agent takes longer than this timeout to reregister, the master shuts down the agent, which in turn will shutdown any live executors/tasks.

Therefore, it is highly recommended to automate the process of restarting an agent, e.g. using a process supervisor such as monit or systemd.

Known issues with systemd and process lifetime

There is a known issue when using systemd to launch the mesos-agent. A description of the problem can be found in MESOS-3425 and all relevant work can be tracked in the epic MESOS-3007.

This problem was fixed in Mesos 0.25.0 for the mesos containerizer when cgroups isolation is enabled. Further fixes for the posix isolators and docker containerizer are available in 0.25.1, 0.26.1, 0.27.1, and 0.28.0.

It is recommended that you use the default KillMode for systemd processes, which is control-group, which kills all child processes when the agent stops. This ensures that "side-car" processes such as the fetcher and perf are terminated alongside the agent. The systemd patches for Mesos explicitly move executors and their children into a separate systemd slice, dissociating their lifetime from the agent. This ensures the executors survive agent restarts.

The following excerpt of a systemd unit configuration file shows how to set the flag explicitly:

[Service]
ExecStart=/usr/bin/mesos-agent
KillMode=control-cgroup