Home > Storage > PowerScale (Isilon) > Product Documentation > Storage (general) > PowerScale OneFS: Cluster Composition, Quorum, and Group State > OneFS 9.1 enhancements
OneFS 9.1 and later include a fast reboot feature that reduces the group change time for planned shutdowns. This is accomplished by reducing the size of the lock cache on all nodes before a node splits or merges to the GMP group. Fast reboot is automatically enabled in OneFS 9.1 and later, with no command line configuration or sysctl required to activate the feature. Fast reboot will apply on upgrade away from OneFS 9.1 to a later release.
The fast reboot feature pre-stages the group change lock handling activity, for example during a planned shutdown. Rather than performing a lock dance during the group change window, the lazy lock queue is instead drained for up to five minutes during the shutdown window. By moving this impactful activity to earlier in the process, it directly benefits OneFS upgrades by reducing the time period of the group change that is disruptive to clients. This directly benefits OneFS upgrades by shrinking the client I/O unavailability window due to group changes to about one second for an average size cluster (around fifteen nodes and up to five million locks per domain).
A two new elements have been added to the group sysctl output to indicate which, if any, nodes have activated the reboot service and are draining locks. For example:
# sysctl efs.gmp.group
efs.gmp.group: <35baa7>(3):{ 1-3:0-5, nfs: 3, isi_cbind_d: 1-3, lsass: 1-3, draining: 1, reboot: 1}
Here, node 1 has activated the reboot service and is draining locks.
Also, the following sysctl output will confirm whether locks are being drained.
# sysctl efs.lk.lki_draining
efs.lk.lki_draining: 0
In this example, a value of zero indicates that no lock draining activity is occurring.