Chapter 2. Introduction to GPFS 43
2.2.5 GPFS management functions availability
In addition to the token requests and the granularity of byte range locking
described in the previous section, another mechanism is needed to check
against failure of components (network adapters, disks, nodes — hardware and
software). This instance checks the conditions of all nodes within the cluster,
which is done within GPFS by the configuration manager (CfgMgr). The CfgMgr
uses two different methods for this purpose:
Disk leasing (request from the node to the CfgMgr)
Pings (request from the CfgMgr to the node)
Both methods are using IP communication to check if the nodes within the cluster
are alive and responsive. Under normal operations the only messages ...