[cgl_discussion] Cluster synchronized hotswap - usage case

Awad, Majid majid.awad at intel.com
Fri Apr 15 16:26:53 PDT 2005

Here is the usage scenario for the Cluster Synchronized device hotswap.
I've checked in the formatted version of this email for archiving.  I'll
be sending another usage case for the forced unmount next week.

CSM.5.0 Cluster Synchronized device hotswap

A description of the requirement (from the requirement document) with
any clarifying text that is appropriate. 

OSDL CGL specifies that carrier grade Linux must have the ability to
perform online insertion and/or online deletion (hot swap) of shared
devices in a cluster.  A hot swap indicator, or device light, should not
be activated when a shared device is prepared for removal until all
operating systems that have access to the shared device have removed
references to the device.  Each node in the cluster must maintain a
consistent cluster-wide view of shared devices.

Example architectures for cluster-synchronized device hot swap are
Advanced Telecom Computing Architecture (ATCA) and Compact Peripheral
Component Interface (cPCI).

Desired Outcome:
Define software architecture and develop a POC implementation for this

System Administrators  

During product upgrades, and routine maintenance.

Implementation Notes:
A storage blade is connected to two compute blades that form a cluster.
The compute blades share the storage on the storage blade with the help
of a global file system like GFS. Now when the storage blade has to be
removed for repair/maintenance the operator opens the latch on the
blade, waits for a visual indication (like a led to stop blinking) and
then removes the blade. The hot swap of the blade will generally be
controlled by a shelf or chassis manager that ensures that it is safe to
proceed with the hot swap operation (based on feedback from the compute
blades) before commanding the blade to cease the visual prompt (stop the
blinking led) and turn off power. Before the compute blades can signal
their hot swap readiness, they will have to cease using the storage on
the blade (so applications will force unmount filesystem, and block
device removal procedures will be completed) to enable graceful
shutdown. Cluster synchronized hotswap feature will provide a cluster
aware mechanism to adequately prepare cluster nodes to handle the
storage device hotswap, before the hotswap operation is completed. 

The implementation of this feature though dependent on an underlying
clustering framework should provide standard interfaces to facilitate
integration with different clustering products. 

More information about the cgl_discussion mailing list