Cluster system and restoration method that performs failover control
A cluster system including a plurality of nodes, a plurality of clusters included in each node and a management module managing the cluster system and an arithmetic module, which are included in each of the clusters, wherein, among all the management modules included in the cluster system, one management module is set representative management module, in the individual clusters, one is set as a master management module, and another is set as a standby management module. Each of the management modules includes a failure monitoring unit and a failover control unit. When a failure in the representative management module is detected by any of the failure monitoring units, any of the management modules included in the non-representative management modules, is set as a new representative management module. A recovery unit restores the failure monitoring unit and the failover control unit in the management module in which a failure is detected.
1. A cluster system including a plurality of nodes, the cluster system comprising:
a plurality of clusters included in each of the nodes,
wherein each cluster includes a management module managing the cluster system and an arithmetic module constituting an arithmetic system,
wherein, among all the management modules included in the cluster system, one of the management modules is set as a representative management module,
wherein in individual clusters of the plurality of clusters, one of the management modules is set as a master management module, and another one of the management modules is set as a standby management module, and
wherein each of the management modules includes:
a failure monitoring unit that monitors an occurrence of a failure in another management module;
a failover control unit that sets, when a failure in the representative management module is detected by any of the failure monitoring units, any of the management modules among non-representative management modules, which are management modules excluding the representative management module among all the management modules, as a new representative management module; and
a recovery unit that restores the failure monitoring unit and the failover control unit in the management module in which a failure is detected by any of the failure monitoring units.
2. The cluster system according to claim 1 ,
wherein each of the management modules has an identifier set for communication,
wherein the representative management module has a predetermined special identifier, and
wherein the failover control unit sets the predetermined special identifier for the new representative management module.
3. The cluster system according to claim 1 ,
wherein the management module and the arithmetic module are realized by different physical processors.
4. The cluster system according to claim 1 ,
wherein the failure monitoring unit monitors the another management module in an identical node, and
wherein when the failure monitoring unit detects a failure in the representative management module, the failure monitoring unit transmits a notification to the failover control unit in the master management module in another of the nodes to cause the master management module to be a new representative management module.
5. The cluster system according to claim 1 ,
wherein each of the management modules has an operating system,
wherein the failover control unit and the failure monitoring unit are realized by executable instructions that operate on the operating system, and the recovery unit operates independently of the operating system, and
wherein the recovery unit further restores the operating system.
6. The cluster system according to claim 1 ,
wherein the arithmetic module includes a storage device and an arithmetic unit that executes an arithmetic operation relating to information recorded in the storage device, and
wherein the arithmetic system is a storage system that records information and reads information.
7. A restoration method performed by a cluster system including a plurality of nodes, the cluster system including a plurality of clusters included in each of the nodes, and each cluster includes a management module managing the cluster system and a storage module constituting an arithmetic system, wherein, among all the management modules included in the cluster system, one of the management modules is set as a representative management module, and wherein in individual clusters of the plurality of clusters, one of the management modules is set as a master management module, and another one of the management modules is set as a standby management module,
the restoration method comprising causing each of the management modules to:
perform a failure monitoring process for monitoring an occurrence of a failure in another management module;
perform a failover process for setting, when a failure in the representative management module is detected by the failure monitoring process in any of the management modules, any of the management modules among non-representative management modules, which are management modules excluding the representative management module among all the management modules, as a new representative management module; and
perform a recovery process for restoring a program that realizes the failure monitoring process and the failover process in the management module in which a failure is detected by the failure monitoring process in any of the management modules.