Techniques to manage non-disruptive san availability in a partitioned cluster

Abstract

Techniques to managing non-disruptive SAN availability in a partitioned cluster comprising one or more components configured to determine whether to separate a cluster into two or more partitions, notify one or more responsive cluster nodes to separate the cluster into a first partition and a second partition, update one or more access states, notify the host that access states of one or more network paths has been updated, and provide the one or more access states. Other embodiments are described and claimed.

1 . A computer-implemented method, comprising:
 providing access states for at least two logical interfaces to a host, the at least two logical interfaces comprising a first and a second logical interfaces, each logical interface being configured to provide access to a first and a second storage units;   receiving status information indicating a partition event has occurred;   transitioning to a majority partition after the partition event;   disassociating from an out of partition cluster node that transitioned to a minority partition after the partition event;   updating at least one access state for at least one logical interface;   notifying the host of the at least updated access state; and   providing the at least one updated access state for the at least one logical interface, such that the host maintains access to the first storage unit via the first logical interface and the second storage unit via the second logical interface after the partition event.   
     
     
         2 . The computer-implemented method of  claim 1 , further comprising:
 determining whether the partition event has occurred based at least partially on the connectivity with a cluster node; and   notifying at least one cluster node with status information for the out of partition cluster node.   
     
     
         3 . The computer-implemented method of  claim 2 , wherein the step of determining further comprises:
 sending at least one heartbeat signal to the cluster node;   determining occurrence of the partition event in the cluster based at least partially on a threshold number of heartbeat response signals received from the cluster node; and   declaring the cluster node as the out of partition cluster node after the cluster node fails to respond to the threshold number of heartbeat signals.   
     
     
         4 . The computer-implemented method of  claim 2 , wherein the status information comprises quorum state information indicating the out of partition cluster node as out-of-quorum and connectivity state information indicating that the out of partition cluster node is out of partition. 
     
     
         5 . The computer-implemented method of  claim 1 , wherein the step of disassociating further comprises:
 evicting at least one component of the out of partition cluster node that transitioned to the minority partition from at least one distributed group; and   leaving at least one distributed group managing at least one storage unit of the out of partition cluster node that transitioned to the minority partition.   
     
     
         6 . The computer-implemented method of  claim 1 , wherein the host is communicatively coupled to the at least two logical interfaces to access the first and second storage units managed by at least two cluster nodes in a cluster, and the partition event separates the cluster into at least two partitions comprising at least the majority partition and the minority partition. 
     
     
         7 . The computer-implemented method of  claim 1 , wherein the at least one access state indicates that at least one storage unit managed by the out of partition cluster node in the minority partition is inaccessible via the at least one logical interface. 
     
     
         8 . A system, comprising:
 a processor circuit; and   memory operatively coupled to the processor circuit, the memory comprising a storage operating system for execution by the processor circuit and at least one component to
 provide access states for at least two logical interfaces to a host, the at least two logical interfaces comprising a first and a second logical interfaces, each logical interface being configured to provide access to a first and a second storage units, 
 transition to a majority partition or a minority partition after a partition event, 
 disassociate from an out of partition cluster node that transitioned to a different partition after the partition event, 
 update at least one access state for at least one logical interface, 
 notify the host of the at least updated access state, and 
 provide the at least one updated access state for the at least one logical interface, such that the host maintains access to the first storage unit via the first logical interface and the second storage unit via the second logical interface after the partition event. 
   
     
     
         9 . The system of  claim 8 , wherein the at least one component is to further
 determine whether the partition event has occurred,   notify at least one cluster node with status information of an out of partition cluster node, and   receive status information indicating a partition event has occurred.   
     
     
         10 . The system of  claim 9 , wherein the at least one component determines whether the partition event has occurred is to further
 send at least one heartbeat signal to a cluster node,   determine occurrence of the partition event based at least partially on a threshold number of heartbeat response signals received from the cluster node, and   declare the cluster node as the out of partition cluster node after the cluster node fails to respond to the threshold number of heartbeat signals.   
     
     
         11 . The system of  claim 9 , wherein the status information comprises quorum state information indicating the out of partition cluster node that transitioned to a different partition as out-of-quorum and connectivity state information indicating that the out of partition cluster node as out of partition. 
     
     
         12 . The system of  claim 8 , wherein the at least component is to further
 evict at least one component of the out of partition cluster node that transitioned to the different partition from at least one distributed group, and   leave at least one distributed group managing at least one storage unit of the out of partition cluster node that transitioned to the different partition.   
     
     
         13 . The system of  claim 8 , wherein the host is communicatively coupled to the at least two logical interfaces to access the first and second storage units managed by at least two cluster nodes in a cluster, and the partition event separates the cluster into at least two partitions comprising at least the majority partition and the minority partition. 
     
     
         14 . The system of  claim 8 , wherein the at least one access state indicates that at least one storage unit managed by the out of partition cluster node in the different partition is inaccessible via the at least one logical interface. 
     
     
         15 . At least one computer-readable storage medium comprising instructions that, when executed, cause a system to:
 provide access states for at least two logical interfaces to a host, the at least two logical interfaces comprising a first and a second logical interfaces, each logical interface being configured to provide access to a first and a second storage units;   transition to a minority partition after a partition event;   disassociate from an out of partition cluster node that transitioned to a majority partition after the partition event;   update at least one access state for at least one logical interface;   notify the host of the at least updated access state; and   provide the at least one updated access state for the at least one logical interface, such that the host maintains access to the first storage unit via the first logical interface and the second storage unit via the second logical interface after the partition event.   
     
     
         16 . The at least one computer-readable storage medium of  claim 15 , wherein the instructions, when executed, further cause the system to:
 determine whether the partition event has occurred based at least partially on the connectivity with a cluster node;   notify at least one cluster node with status information for the out of partition cluster node; and   receive status information indicating the partition event has occurred.   
     
     
         17 . The at least one computer-readable storage medium of  claim 16 , wherein the instructions to determine whether the partition event has occurred, when executed, further cause the system to:
 send at least one heartbeat signal to the cluster node;   determine occurrence of the partition event based at least partially on a threshold number of heartbeat response signals received from the cluster node; and   declare the cluster node as the out of partition cluster node after the cluster node fails to respond to the threshold number of heartbeat signals.   
     
     
         18 . The at least one computer-readable storage medium of  claim 15 , wherein the instructions to disassociate from the out of partition cluster node, when executed, further cause the system to:
 evict at least one component of the out of partition cluster node that transitioned to the majority partition from at least one distributed group; and   leave at least one distributed group managing at least one storage unit of the out of partition cluster node that transitioned to majority partition.   
     
     
         19 . The at least one computer-readable storage medium of  claim 15 , wherein the host is communicatively coupled to the at least two logical interfaces to access the first and second storage units managed by at least two cluster nodes in a cluster, and the partition event separates the cluster into at least two partitions comprising at least the majority partition and the minority partition. 
     
     
         20 . The at least one computer-readable storage medium of  claim 15 , wherein the at least one access state indicates that at least one storage unit managed by the out of partition cluster node in the majority partition is inaccessible via the at least one logical interface.