12-03-2013 06:51 PM
Hello,
We are planning to implement global clustering between 2 sites. We would like to run apache on the server, if there is any issue with one of the site, we need to run it from other site.
I'm trying to understand the concept and I read the document.
hearbeat IP address and CluserIP address. Do we need to use the same virtual IP address for heartbeat and clusterIP ?
Solved! Go to Solution.
12-03-2013 07:48 PM
Hello,
WAC resource is an ICMP connection between your two sites which works on IP address. ICMP packets are sent between two sites to check the "ALIVE" status of DR site.
Couple of bullet points to remember
Each VCS cluster must have a valid virtual IP address to be used as the clusters ‘ClusterAddress’ for ICMP and WAC communication. This address should be made highly available such that it is always configured on one node in the VCS cluster (normally by adding the address to the ClusterService service group)
Each VCS clusters ‘ClusterAddress’ should be resolveable to a host name (i.e. using DNS for example)
You can use "Add cluster" wizard to configure global cluster, this will require the ClusterAddress IP address of all clusters being added to the GCO cluster.
an e.g main.cf is like
remotecluster test20001 (
ClusterAddress = "10.12.251.191"
)
heartbeat Icmp (
ClusterList = { test20001 }
AYATimeout = 30
Arguments @test20001 = { "10.12.251.191" }
)
Hope this helps
G
12-04-2013 01:27 AM
Yes heartbeat IP address and ClusterIP address are normally the same so cluster ips should be set in 4 places:
I think technically you could use a separate IP for the heartbeat, but I can't think of a reason why you would want to do this. If you have a separate independent link across sites, then I believe you can add a second global cluster heartbeat if you wanted
I believe cluster information is transfered over the cluster IP addres and heartbeats over the heartbeat address, which, as above, are normally the same IP, so provide the same functionallity as GAB, but is completely separate from GAB.
Mike
12-04-2013 08:46 AM
Hello,
Yes you are right, if an incorrect IP is assigned or somehow the two cluster IP addresses can't communicate to each other, the status of remote cluster won't be known to cluster. Ideal state of heartbeat will be displayed as "Alive" in hastatus -sum output which won't be case in afore said situation.
About the resolution of hostname to IP, I don't believe its mandatory but its good to have as it can help to mark easy names for server , sites or IP addresses .. Only putting IP addresses will make it difficult to remember.
G
12-03-2013 06:57 PM
Also, WAC is transfer information between cluster. Does it do same functionality as GAB doing between the node?
12-03-2013 07:48 PM
Hello,
WAC resource is an ICMP connection between your two sites which works on IP address. ICMP packets are sent between two sites to check the "ALIVE" status of DR site.
Couple of bullet points to remember
Each VCS cluster must have a valid virtual IP address to be used as the clusters ‘ClusterAddress’ for ICMP and WAC communication. This address should be made highly available such that it is always configured on one node in the VCS cluster (normally by adding the address to the ClusterService service group)
Each VCS clusters ‘ClusterAddress’ should be resolveable to a host name (i.e. using DNS for example)
You can use "Add cluster" wizard to configure global cluster, this will require the ClusterAddress IP address of all clusters being added to the GCO cluster.
an e.g main.cf is like
remotecluster test20001 (
ClusterAddress = "10.12.251.191"
)
heartbeat Icmp (
ClusterList = { test20001 }
AYATimeout = 30
Arguments @test20001 = { "10.12.251.191" }
)
Hope this helps
G
12-04-2013 01:27 AM
Yes heartbeat IP address and ClusterIP address are normally the same so cluster ips should be set in 4 places:
I think technically you could use a separate IP for the heartbeat, but I can't think of a reason why you would want to do this. If you have a separate independent link across sites, then I believe you can add a second global cluster heartbeat if you wanted
I believe cluster information is transfered over the cluster IP addres and heartbeats over the heartbeat address, which, as above, are normally the same IP, so provide the same functionallity as GAB, but is completely separate from GAB.
Mike
12-04-2013 04:53 AM
Thank you very much Mike and Gurav.
IF I understand correctly, WAC process uses ClusterIP address and ICMP to check the cluster hearbeat and cluster information. If we have wron IP address on cluster information and if I want to find the remote resources/service group states, would it say as unknow state or offline?
12-04-2013 04:57 AM
Also, why IP address hould be resolve to hostname? Is it importnate to have on DNS?
12-04-2013 08:46 AM
Hello,
Yes you are right, if an incorrect IP is assigned or somehow the two cluster IP addresses can't communicate to each other, the status of remote cluster won't be known to cluster. Ideal state of heartbeat will be displayed as "Alive" in hastatus -sum output which won't be case in afore said situation.
About the resolution of hostname to IP, I don't believe its mandatory but its good to have as it can help to mark easy names for server , sites or IP addresses .. Only putting IP addresses will make it difficult to remember.
G
12-04-2013 09:00 AM
Thank you very much for all the info.
12-04-2013 03:31 PM
FYI - the cluster IP must be in DNS (partly for the reasons Gaurav mentioned above; is it also specifically mentioned as a requirement in the documentation)
Prerequisites for global clusters -> Prerequisites for cluster setup
Every cluster must have a valid virtual IP address, which is tied to the cluster. Define this IP address in the cluster's ClusterAddress attribute. This address is normally configured as part of the initial VCS installation. The IP address must have a DNS entry.
12-04-2013 06:58 PM
Thank you very much, but why DNS is Prerequisite?
12-04-2013 07:45 PM
short answer: ease of management ie: the reasons Gaurav pointed out
it's obviously been listed in the document for a reason (ie: it won't work properly/won't have full functionality without it) - if you're really hell bent on challenging why it's required, log a case with support.