Network and Storage Protocols

Brocade warm reboot

Alexey1
866 Views

Hi there

when we do a FabricOS upgarde for a "pizza box" switch, at some stage Fabric OS does a warm reboot to boot a new version of software.

At the same time the traffic is continued to being served. I wonder how exactly is it implemented. Could not find any documentation.

Have you seen any detialed explanations, some tech articles?

 

Regards,

Alexey

7 REPLIES 7

cedric_renauld
813 Views

In case of SAN Brocade, normaly, you may have 2 Fabric, a minimm of tow switches, 1 for Fabric 1 and another for Fabric 2

This multiple fabric give to your hosts a multipath access to your LUN storage

When you update a SAN fabric, the IO are frozen, but you can do that fabric after fabric, that give to your hosts always one path to the storage

Alexey1
780 Views

Well, it's not about having two fabrics. According to the FOS Upgrade Guide, the warm reboot is nondisruptive.  Only Generic Services are affected, so the switch will not be able to register a new device for example.

However, switch continues to transfer frames. So IO is not frozen.

Im curious how this is achieved.

 

 

All Brocade systems maintain two partitions (a primary and a secondary) of a nonvolatile storage to store the firmware.
The firmware download process first copies the replacement files (containing an updated kernel) into the secondary
partition. Then, the process swaps the partitions so that the secondary partition becomes the primary. It then performs
a nondisruptive HA reboot of the system. For directors, the standby is rebooted; this does not affect the system traffic. The
system attempts to restore the previous machine state after the reboot is completed for fixed-port platforms, also called a
warm reboot. When the system boots, it boots using the revised Fabric OS firmware in the primary partition. The firmware
download process then copies the updated files from the primary partition to the secondary partition.
NOTE
Most firmware upgrades and downgrades are not disruptive to device operations; however, always refer to the
latest Fabric OS release notes for updates on upgrading and downgrading.

 

NOTE
During the (HA) reboot on fixed-port switches, exchanges involving Fibre Channel Generic Services may
experience a delay. Fixed-port switches must retry the operations in this case.

 

 

Vijay5
613 Views

Hi Alexey,

 

To perform warm reboot, kindly use the command hareboot.

 

Use this command to perform warm reboot in the switch-based systems. On chassis-based systems, this command
behaves similar to the haFailover command so that the standby CP becomes the active CP.
In chassis-based systems, if the active and standby CPs are not synchronized or the system is not in the redundant mode,
the command aborts.
When the switch-based system is not fully up or stable, the haReboot command cannot be completed.

 

Ref:

https://techdocs.broadcom.com/content/dam/broadcom/techdocs/us/en/pdf/fc-networking/software-fabric-os/fos-92x-command.pdf

Alexey1
602 Views

This doesn't respond to my question at all, unfortunatelly. 

pmidhun
471 Views

Normally the clients/hosts will be having a physical multipath/redundant connection - same host connects to both the switches and has effective zoning available on both.  So even when one switch is unavailable, you are still connected to the redundant switch and data traffic continues to flow over it. Something like the below.

pmidhun_0-1730266563140.jpeg

Does this answer your query?

Alexey1
459 Views

Yes, sure, there are two fabrics which are redundant but the point is not about a fabric being unavailable.

As it's stated in the brocade documentation, the traffic is not disrupted during the warm reboot of a switch. the frames continue to flow. only the generic fabric services are unavailable during this operation. that is the object of my interrogation. how is it achieved. 

Vijay5
451 Views
Hi Alexey,

There will be two partitions in the switches, primary and secondary partition.
During the reboot secondary partition will get upgraded first and followed by the primary partition.
It will be the same way when we perform the HA reboot, I could not find any brocade document which explains the same.

Regards,
Vijay
Public