- play_arrow Overview
- play_arrow Storage Overview
-
- play_arrow Data Center Bridging (DCBX, PFC)
- play_arrow Using Data Center Bridging (DCBX, PFC)
- Understanding DCB Features and Requirements
- Understanding DCBX
- Configuring the DCBX Mode
- Configuring DCBX Autonegotiation
- Disabling the ETS Recommendation TLV
- Understanding DCBX Application Protocol TLV Exchange
- Defining an Application for DCBX Application Protocol TLV Exchange
- Configuring an Application Map for DCBX Application Protocol TLV Exchange
- Applying an Application Map to an Interface for DCBX Application Protocol TLV Exchange
- Example: Configuring DCBX Application Protocol TLV Exchange
- Understanding CoS Flow Control (Ethernet PAUSE and PFC)
- Example: Configuring CoS PFC for FCoE Traffic
- play_arrow Learn About Technology
-
- play_arrow Configuration Statements and Operational Commands
Understanding FCoE and FIP Session High Availability
High availability features maintain storage network sessions when a system process is terminated and during certain types of upgrades:
High Availability for Fibre Channel Process Termination (FCoE-FC Gateway Mode, QFX3500 Only)
In FCoE-FC gateway mode, the QFX3500 switch provides high availability to restore the FCoE sessions running on the switch in case the Fibre Channel (FC) process is terminated. A session is a fabric login (FLOGI) or fabric discovery (FDISC) login to the FC SAN fabric, not an end-to-end server-to-storage session.
The switch stores FCoE session data in a persistent storage module. If the FC process terminates, the switch restores the existing FCoE sessions on the same interfaces that they were on before the FC process terminated. Data traffic for existing sessions is not affected during session restoration.
For a brief time, the system does not process control traffic because of the FC process restart and session restoration. During this brief time, no new FCoE sessions can be established, and no existing sessions can log out.
During the restoration process, if the FC process does not receive an interface up notification from a particular interface within a certain time, the switch times out the restore operation and discards the data on that interface. The previously existing FCoE sessions on that interface are not restored, and the ENodes must log in again.
An FC process restart and session restoration resets the Fibre Channel statistics.
If the FC process terminates repeatedly, the operating system
disables the process until you manually restart it. To restart the
FC process manually, issue the restart fibre-channel command.
High Availability for FIP Snooping
You can configure the system to perform FIP snooping on Ethernet interfaces that are connected to FCoE devices that have ENodes. The high availability function restores running FIP snooping sessions in case the Ethernet switching process is terminated.
QFX10000 switches do not support FIP snooping. You don’t need to enable FIP snooping on aggregation devices because FIP snooping is performed at the FCoE access edge.
The Ethernet switching process stores the FIP snooping state in a persistent storage module. If the Ethernet switching process terminates, the switch restores the existing FIP snooping sessions on the same interfaces that they were on before the Ethernet switching process terminated. The high availability features preserve:
Logged in ENodes
Discovered FCFs
Existing sessions
Existing FIP snooping filters
The complete restoration process, including reconciling all valid states, takes a maximum of 8 seconds. During the restoration process, the switch can learn a new FCF or a new FC switch, and new ENodes can log in to the FC network. However, FDISC messages from an ENode that is already logged in to the network might be dropped if the ENode has not yet been restored.
When the Ethernet switching process terminates ungracefully, the FIP keepalive timer is reset to the normal initial value, not the value at the time of the Ethernet switching process termination.
In the event of an Ethernet switching process termination, ENodes remain logged in, and existing sessions are not interrupted.
An Ethernet switching process restart and session restoration resets the FIP snooping statistics.
Nonstop Software Upgrade (QFabric Systems)
On QFabric system Node groups that have more than one Node device, nonstop software upgrade (NSSU) enables you to upgrade the Node devices with minimal packet loss and maximum uptime. NSSU automates software upgrades on the QFabric system components in an orderly and consistent manner to maximize system uptime.
The system upgrades components with redundant architectures, such as redundant server Node groups and network Node groups that have two or more members, in stages. While the system upgrades one component, the redundant component continues to function.
For example, while one member of a redundant server Node group is upgraded, the other member continues to forward traffic. When the first Node group member completes the upgrade, it comes online while the system upgrades the second member.
NSSU provides high availability for the lossless traffic forwarding required to support storage networks. If your system design includes redundancy (redundant Node devices in Node groups, LAGs, and so on) so that an alternate traffic path is available, when you upgrade a Node device, traffic is not impacted.
In fully redundant topologies, NSSU preserves FIP session, FIP snooping filter, VN2VF_Port session, and VN2VN_Port session information and prevents traffic loss in most cases. An exception is that Node devices that are directly connected to ENodes experience momentary traffic loss when the Node device reboots.
