Hot Spare Nodes. Hot Spare Node (HSN) feature available from Spectrum Virtualize version 8.1.0 and later, enables users to configure Hot Spare Nodes to their existing SVC cluster. These Spare Nodes can automatically swap the failed nodes in cluster, thereby keeping the system up during the software upgrades, during maintenance or during unplanned outages.
Enablement of Hot Spare Node functionality provides umpteen benefits with respect to smooth upgrade experience, hardware redundancy, more effective multi-path coverage etc.
When a SVC node goes for software upgrade, it's peer node in IO group writes through the cache, thereby handling both node's workload. With HSN enabled, cluster will take a Spare Node in place of the node that has gone down for upgrade, therefore the cache doesn't need to go write through. Once the upgraded node returns, Spare node goes back, without any failover and failback seen at the multipathing layer at host side.
All of this is handled by the NPIV ports, so no more nervous times during upgrades!
Let’s talk a bit about NPIV here.
NPIV (N-Port ID Virtualization) is a standard that describes how a single fibre channel port register with SAN fabric using multiple WWPNs, termed as virtual WWPNs.
Spectrum Virtualize version 7.7.0 and later enabled NPIV support, that allow users to use separate ports for Back-end communication, Inter-node communication and for host communication. Use of NPIV target ports improves the host experience by separating out host communications from Inter-node and Back-end communication.
Three transition states or “fctargetportmode” states are provided for the seamless enablement of NPIV feature in existing as well as new SVC configurations.
When a node of an IO group goes down, it’s partner node’s fail-over ports become active and failed nodes NPIV ports migrate over it. That means Spectrum Virtualize software will perform an NPIV fail-over. Hot Spare Node functionality is based on fail-over of these NPIV Ports.
Let's see how Hot Spare Node functionality works.
A Spare node is added to the cluster, in the same way as other nodes in cluster, it essentially becomes another node in cluster with identity as “Spare”, but not part of any IO group.
In the event of any node failure, Spare will use the NPIV nature of the host virtual ports to take over the personality of the failed node and becomes active with status as “Online_Spare”. As we have NPIV enabled, the host won’t see any missing paths in this duration.
Here is a diagrammatic representation of NPIV and Hot Spare Node Functionality.
Figure 1: Two Nodes in SVC IO group. Primary Ports are used for Clustering and Backend communication. NPIV ports used for Host communication.
Figure 2: One of the Node of IO group fails, it's NPIV ports failover to peer node.
Figure 3: Failed node's NPIV ports failover to Spare node in Cluster.
Failover of the Spare nodes is automatic, however manual swapping of Spare nodes can be done using “Swapnode” CLI.
Maximum 4 Spare nodes can be added to the cluster. So, the maximum configuration can have 12 SVC nodes in cluster, i.e. 8 Active nodes and 4 Spare nodes. These Spare nodes protect the active nodes, based on matching hardware type and configuration of active cluster nodes. The list of nodes which a Spare protects is listed in “protected nodes” field in the node view.
The Spare nodes are also available as an option on the FlashSystem V9000 that use an external switch for Flash enclosure attachment.
Spectrum Virtualize clusters can be split across multiple data centers, like in Enhanced Stretched cluster or in Hyper-Swap configurations.
This feature has been in wish list for SVC customers for long time and now with HSN enabled, users are ready to explore umpteen benefits of this feature and get even more power and redundancy in their SVC configurations.
For more details and deep dive of this feature, you may refer Redpaper:
IBM Spectrum Virtualize: Hot-Spare Node and NPIV Target PortsThank You.
#Softwaredefinedstorage#IBMSoftwareDefinedStorage#Flashstorage
