• No results found

Starting the Cluster Services on the Master Node

MCS supports both multicast and unicast for intra-cluster communication. This body of this guide provides instructions for configuring a cluster in a multicast environment (standard configuration). However, multicast requires multicast enabled routers. If your network does not support multicasting, see Unicast Support in Clustering for details on altering the configuration.

1. On the Master node only, assign the cluster multicast IP address. This is the IP that the cluster will use for communication between the nodes.

a. If your network has no other multicast activity, the default multicast address of 239.192.1.1 is assumed in the following command:

/opt/avid/cluster/bin/cluster setup-corosync --corosync-bind-iface=interface --rabbitmq_master="master hostname"

b. If your network includes multicast activity (perhaps a second MCS system), specify a custom multicast address with the following command:

/opt/avid/cluster/bin/cluster setup-corosync --corosync-bind-iface=interface --corosync-mcast-addr="multicast address" --rabbitmq_master="master hostname"

Note: As of MCS v2.0, this command will not accept a multicast address outside of the 239.0.0.0/8 range. If specifying a non-239.x.x.x, you will see an error:

“Multicast IP is not in subnetwork 239.0.0.0/8”.

Explanation (do not type this example)

This procedure uses the cluster setup-corosync command:

cluster setup-corosync

[corosync-bind-iface =<interface>]

{[--corosync-mcast-addr="<ip>"]}

{[--rabbitmq_master="<device>"]}

where:

--corosync-bind-iface: Identifies the name of the primary network interface. In an HP server, this is generally “eth0”. In a Dell server, this is generally “em1” for 1 Gb connections or “p1p1” / “p2p1” for 10 Gb connections (depending on which slot the card 10 Gb card has been installed). For Interplay MAM configurations with port bonding, this is generally “bond0”. Quotes are not required in this command.

--corosync-mcast-addr: In configurations that do not use the default multicast address of 239.192.1.1, this command can be used to specify a custom multicast address. Quotes are required in this command.

--rabbitmq_master: This specifies the (short) hostname of the Master node in the cluster (e.g. wavd-mcs01). This should be the same as the DRBD Master node

Messages appear echoing the Corosync network binding process; followed by messages indicating that services are being stopped. At the end of the process, you are informed that the Corosync cluster engine has successfully started [OK].

The following is sample output:

bind_iface=eth0 bind_network=192.168.10.51 mcast_addr=239.192.1.1 Avid Service: edit fl_xmd: no process killed

. ..

...

Starting Corosync Cluster Engine (corosync): [ OK ] Starting Pacemaker Cluster Manager: [ OK ] You may notice the following text appear during this process:

Clustering node rabbit@nodename with rabbit@nodename...

Error: cannot_cluster_node_with_itself Starting node rabbit@nodename....

…done

Failed to join cluster, exiting!!.

This message can be ignored as it simply indicates that this is the first node in the rabbit cluster.

2. On the Master node only, assign the cluster’s virtual unicast IP address. This is the IP that the cluster will use for communication with clients and external systems.

/opt/avid/cluster/bin/cluster setup-cluster --cluster_ip="cluster IP address" pingable_ip="router IP address"

--cluster_ip_iface="eth0" --admin_email="comma separated e-mail list" --drbd_exclude="comma separated list of non-DRBD nodes"

Explanation (do not type this example)

This procedure uses the cluster setup-cluster command:

cluster setup-cluster

{[--cluster_ip="<cluster IP address>"]}

{[--pingable_ip="<router IP address>"]}

{[--cluster_ip_iface="<interface_name>"]}

{[--admin_email="<comma separated e-mail list>"}]

{[--drbd_exclude="<comma separated list of non-DRBD nodes>"}]

where:

--cluster_ip: Identifies the unicast virtual IP address assigned to the cluster.

--pingable_ip: This is an IP address that will always be available on the network - for example, the IP address of your default gateway (e.g. 192.168.10.1).

--cluster_ip_iface: Identifies the name of the primary network interface. In an HP server, this is generally “eth0”. In a Dell server, this is generally “em1” for 1 Gb connections or “p1p1” / “p2p1” for 10 Gb connections (depending on which slot the card 10 Gb card has been installed). For Interplay MAM configurations with port bonding, this is generally “bond0”.

--admin_email: This is a comma separated list of e-mail addresses to which cluster status notifications are automatically sent. This command is not optional. If you do not want to receive e-mail notifications, enter a bogus email address.

Note: At least one cluster administrator email address is mandatory (though not validated by the system). To change the email address later, see “Changing the Cluster Administrator Email Address” on page 202.

--drbd_exclude: This is a comma separated list of the non-DRBD nodes in the cluster (e.g. wavd-mcs03, wavd-mcs04). This parameter prevents the non-DRBD nodes from running PostgreSQL. The comma-separated list of non-DRBD nodes must not contain any spaces between each entry, only a comma.

Note: Quotes are required in each of the above commands.

Error messages appear indicating missing resources and attributes.

For example:

ERROR: resource <resource name> does not exist

Error: performing operation: The object/attribute does not exist These can be ignored.

Additional warning, error and info messages may also appear, similar to the following:

WARNING: 125: AvidConnectivityMon: specified timeout 20s for start is smaller than the advised 60

ERROR: 125: rsc-options: attribute admin-email does not exist INFO: 125: commit forced

These can be ignored.

3. Restart the following services so they register correctly on the newly created instance of the message bus:

service avid-acs-messenger restart service avid-aaf-gen restart

4. Now that the clustering services are up and running on the master node, start the cluster monitoring tool:

crm_mon

This utility provides a “live view” of the cluster which can be useful as you add additional nodes to the cluster. Typing “crm_mon –f” will give you additional information about fail counts. Press CTRL-C to exit crm_mon.

Note: If you are using a SSH tool such as PuTTY, open a second session to the Master

Related documents