Migrating Symantec Veritas Cluster Server to HP Serviceguard for Linux

Technical white paper | Migrating Symantec Veritas Cluster Server to HP Serviceguard for Linux

• HP Continentalclusters for Linux: HP Continentalclusters protects and manages multiple clusters where the clusters

are located apart from each other over any distance. With Continentalclusters one can perform a push-button recovery

on the recovery site and use any type of replication to replicate data between the data centers. HP provides preintegrated

solutions with Continentalclusters and disk array based replication for the following arrays like HP 3PAR P10000/P7000

Remote Copy, HP XP P9000 Continuous Access, and EVA P6000 Continuous Access. In addition, you can integrate any

host based replication technology with Continentalclusters using the data-sender and data-receiver framework in

Continentalclusters. You can also use Continentalclusters with other non HP disk-array based replication. In such cases,

Continentalclusters can be used to automate the recovery process, and the user is responsible for data currency and

consistency before initiating a recovery.

SG/LX cluster membership

An SG/LX cluster is a networked grouping of HP ProLiant systems (host systems known as nodes) having sufficient redundancy

of software and hardware that a single failure may not significantly disrupt service. Application services (individual Linux

processes) and their corresponding storage and network resources are deployed together using the package construct. In the

event of a single service, node, network, or other resource failure, SG/LX can automatically transfer control of the package to

another node within the cluster, allowing services to remain available with minimal interruption.

• Cluster quorum

In the case of a failure, SG/LX cluster reforms with the surviving nodes. In SG/LX each cluster node has one vote. During

cluster reformation, the nodes that can communicate with each other, regroup and attempt to reform a cluster. In SG/LX,

a cluster can be reformed only if the cluster quorum condition is achieved. The cluster quorum condition is met only when

a strict majority (that is, more than 50%) of the nodes running previously is available to form the cluster. All and any

remaining nodes that are not in the group with cluster quorum condition can be reset. Since the quorum is dynamic (The

required quorum is generally based on the previously running cluster membership and not on the static count of the

number of nodes configured in the cluster.), SG/LX clusters can reform even after repeated multiple cluster node failures.

• Split brain

If a failure in the SG/LX cluster results in equal halves (exactly 50%) of a previously running cluster, then this situation is

called as split brain. A robust arbitration method is used to break a tie. This robust arbitration method guarantees that

only one half can form a cluster and the other half may be forced to shut down. In SG/LX, arbitration can be implemented

by cluster lock. Either Lock logical unit number (LUN) or Quorum Server can be configured as cluster lock. Configuring a

membership arbitration method (cluster lock) is a must in the clusters that have two nodes.

• Lock LUN

Lock LUN acts as a quorum arbitration method in the SG/LX clusters. It is a special area in the storage which is sharable by

all nodes in the cluster. During split brain situation, each partition races with the other to obtain a lock on the Lock LUN.

Once the lock is obtained, it is marked so that the other partition can recognize the Lock LUN as “taken”. Serviceguard

periodically checks the status of Lock LUN device and logs messages in the syslog in case of a failure. The Lock LUN is

dedicated to use as only membership arbitration method, which is also called as cluster lock. HP recommends that the

Lock LUN should comprise of whole disk. The Lock LUN can be used only in the cluster which has nodes up to four. The

following diagram shows Lock LUN configuration in a 2-node cluster. When there is loss of communication between

node 1 and node 2, both nodes race to acquire the lock and node 2 wins the race by acquiring the lock. The node 2 is

allowed to continue running and node 1 is reset.

Figure 1. 2-node SG/LX cluster using Lock LUN as a cluster lock