Brocade FICON Administrator's Guide v7.1.0 (53-1002753-01, March 2013)
60 FICON Administrator’s Guide
53-1002753-01
Firmware management in a FICON environment
5
Firmware upgrade disruption
Loading new firmware is a non-disruptive process except for the following cases:
• During the firmware download process a failover to the backup processor occurs. Any
outstanding CUP commands are lost and result in a timeout. To avoid this, CUP should be
varied offline before starting a firmware download. The CUP may be varied back online after a
successful firmware download.
• All firmware downloads will cause an I/O disruption when the new code is activated on the
FX8-24 blade and 7800 switch.
NOTE
If upgrading multiple switches or blades, upgrade one at a time. This is not a requirement for DCX
or DCX-4S platforms.
Non-disruptive firmware upload and download
All systems maintain two partitions of nonvolatile storage areas, a primary and a secondary.
Following is an overview of the firmware download process:
1. The firmware is downloaded to the secondary partition.
2. The current secondary and primary partitions are swapped.
3. The CP then reboots. In dual CP systems (Brocade 48000 and Brocade DCX enterprise-class
platforms) the backup CP is rebooted with the new firmware and the HA failover is
automatically invoked.
4. Once rebooted, the primary code image is copied to the secondary.
Considerations for fixed port switches
Unlike director-class products, fixed port switches have only one control processor (CP) so they
behave differently during Fabric OS hot code upgrades. This can impact FICON operations if best
practice is not followed.
During the code upload process, there is a period of time when the CP is busy rebooting the
operating system and re-initializing Fabric OS, and it is not available to respond to CPU-targeted
queries/frames. For dual-CP products, this “busy” time is a couple of seconds (the time it takes for
mastership to change), and is non-disruptive. For single-CP fixed port switches, however, the “busy”
period can last a minute or two. During this time, data traffic will continue to flow uninterrupted,
unless a response from the CP is required. If a response from the CP is required, for path and
device validation for instance, Interface Control Checks (IFCCs) are possible. Any IFCCs should be
recovered automatically by the System z host, with no required user action.
This does not mean that switches are not appropriate for FICON. Fully certified FICON switch
deployments will continue to be sold and supported. However it is important to follow best
practices of performing the firmware upgrade during scheduled maintenance windows where
traffic is minimized, IFCCs can be tolerated, and CUP is turned off.