Control hosts (Linux)

You can start, shut down, and restart local, remote, or multiple Linux hosts. Find information on each of these options.

Start hosts

Log on with root permissions and rsh on all hosts in the cluster. To start a hosts specified by name, you need to be able to run rsh across all hosts in the cluster without having to enter a password; see your operating system documentation for information about configuring rsh.

Important:

By default, only root can start, stop, or restart the cluster. Optionally, you can grant root privileges to egoadmin, the cluster administrator account.

Starting hosts brings them into the cluster where they become usable resources.

Start a local Linux host

Log on with root permissions.

To start a local host, take the following steps.

Start EGO on your local host:

egosh ego start

Start a remote Linux host

Log on with root permissions and rsh on the local host and the remote host. If the master is up, log onto any host in the cluster; if the master is down, log onto a management host.

Start EGO on a remote host:

ego start host_name

Replace host_name with the name of your remote host.

Start multiple Linux hosts

Log on with root permissions and rsh on the local host and each remote host. If the master is up, log onto any host in the cluster; if the master is down, log onto a management host.

Start EGO on multiple hosts:

ego start host_name1 host_name2 ...

Replace host_name with the names of your remote hosts. Separate host names with a space.

Shut down hosts

Shutting down a host immediately changes the host state to Unavailable. Running workload is automatically restarted on another host. If you need to perform maintenance, you can choose to close a host instead of shutting it down.

Stop a local Linux host

Log on with root permissions.

Stop EGO on your local host:

egosh ego shutdown

Stop a remote Linux host

Log on with root permissions on the local host. If the master host is up, log onto any host in the cluster; if the master is down, log onto a management host.

Stop EGO on a remote host:

egosh ego shutdown host_name

Replace host_name with the name of your remote host.

Stop multiple Linux hosts

Log on with root permissions on the local host. If the master host is up, log onto any host in the cluster; if the master is down, log onto a management host.

Stop EGO on multiple hosts:

egosh ego shutdown host_name1 host_name2 ...

Replace host_name with the names of your remote hosts. Separate host names with a space.

Enable automatic expiry of unavailable compute hosts

By default, all hosts that join the cluster remain in the cluster, even if they become unusable.

Optionally, after a period of time in the Unavailable state, a compute host can expire from the cluster—it no longer appears in the Console, or in egosh resource list or resource view output, and it is not affected by egosh start all or restart all commands.

Host expiry is not irreversible. If you restart an expired host (for example, run egosh ego start host_name or restart the expired host while automatic system startup is configured), it can rejoin the cluster in the same way that a newly installed host joins the cluster.

If you want to remove a management host from the cluster, a different procedure is required.

To configure the host expiry feature, take the following steps.

  1. Logon as egoadmin and edit ego.conf.
  2. Add the parameter EGO_DYNAMIC_HOST_TIMEOUT and specify an expiry period.

    The default time period is in hours. Use an M after the time value to represent minutes (for time periods of 10 minutes or more).

    For example:

    • EGO_DYNAMIC_HOST_TIMEOUT=48 means the unavailable host is removed after 48 hours.

    • EGO_DYNAMIC_HOST_TIMEOUT=75M means the host is removed after an hour and fifteen minutes.

    • EGO_DYNAMIC_HOST_TIMEOUT=2M means the host is removed after 10 minutes; a value of 2 minutes is below the allowable 10 minute minimum.

  3. Restart the master host for the change to take effect.

Remove management hosts from the cluster

You must have previously enabled automatic expiry of unavailable compute hosts.

Hosts that have been added to a cluster using the command egoconfig mghost, and are designated as management hosts, need to be physically deleted from ego.cluster.cluster_name if you want to remove them from the cluster. This requirement applies to current management hosts.

  1. Shut down the host.
  2. If you have configured automatic startup during your cluster setup, then run the command egoremoverc.sh as root.

    Doing this prevents automatic startup when the host reboots, and keeps the host from being re-added dynamically to the cluster.

  3. If the host is a master candidate, run the command egoconfig masterlist to remove the host from the failover order.
  4. Run egoconfig unsetmghost to remove the host from the management host group.

    Running this command removes the host entry from ego.cluster.cluster_name.

  5. Optional. If you want to completely remove the host from the cluster, you can shutdown all the applicable services and daemons at this point, and then remove the directory where the Linux installation is located.
  6. Restart the master host to change it from a management host to a compute host.

    Because the host is shut down, and daemons are no longer running, the host switches to an unavailable state. Now that you’ve got a compute host in an unavailable state, complete the steps for Enable automatic expiry of unavailable compute hosts.

Restart hosts

You may want to restart the EGO daemons on a host if it is in an Unavailable state. An unavailable host may have problems with memory or unnecessary applications that can be fixed by restarting it.

Restart a local Linux host

Log on with root permissions.

Restart EGO on your local host:

egosh ego restart

Restart a remote Linux host

Log on with root permissions. If the master host is up, log onto any host in the cluster. If the master is down, log onto a management host.

Restart EGO on a remote host:

egosh ego restart host_name

Replace host_name with the name of your remote host.

Restart multiple Linux hosts

Log on with root permissions. If the master host is up, log onto any host in the cluster; if the master is down, log onto a management host.

Restart EGO on multiple hosts:

egosh ego restart host_name1 host_name2 ...

Replace host_name with the names of your remote hosts. Separate host names with a space.