Howto: 50Gb tri-speed transceivers
Overview
HPE Aruba Networking recently released a major new firmware upgrade for CX switches with CX 10.12. One of the interesting new features is that the standard 50Gb multimode transceiver has been enabled with tri-speed mode. That's right, if you already have a R0M48A 50Gb transceiver, when you use it in switch with CX 10.12, you can run at 10, 25 or 50Gb speeds!
I will show how to do an upgrade from 10 to 25 to 50Gb with minimal downtime in this post.
Scenario
This could be useful in scenarios such as:
- You upgrade core/aggregation switches first (eg with the CX 8360, CX 6400 models with 50Gb ports) but still need to connect to the 10Gb downstream access switches: put in 50Gb optics and run them at 10Gb until you can upgrade the access switches
- Conversely, the access switches have been replaced with CX6300, and you want to run 25 or 50Gb uplinks, but the upstream core/agg devices only have 10 or 25Gb ports still: include 50Gb optics in the 6300 access switches and run them at the lower speed until you can upgrade the core/agg switches
Example Step Upgrade Process
Environment
- 2 x 6300 24 port switches
- The upper switch is 6300-aggregation
- The lower switch is 6300-access
- Both switches have a pair of SFP56 ports configured as an LACP LAG (1/1/27 & 1/1/28)
In the starting configuration, the aggregation switch has an LACP LAG group running over a pair of 50Gb tri-speed transceivers configured to run at 10Gb, connected to the access switch with a pair of 10Gb transceivers.
Aggregation Switch Config
interface lag 1
description Downstream Lag
no shutdown
no routing
vlan trunk native 930
vlan trunk allowed all
lacp mode active
lacp fallback-static
lacp rate fast
exit
interface 1/1/27
no shutdown
speed auto 10g
! actual speed 10g
lag 1
exit
interface 1/1/28
no shutdown
speed auto 10g
! actual speed 10g
lag 1
exit
Access Switch Config
interface lag 1
description Uplink-Lag
no shutdown
no routing
vlan trunk native 930
vlan trunk allowed all
dhcpv4-snooping trust
lacp mode active
lacp rate fast
exit
interface 1/1/27
no shutdown
lag 1
exit
interface 1/1/28
no shutdown
lag 1
exit
Central, MultiEdit and LACP
- If you are managing these switches with Central, it is best to use MultiEdit to make these changes.
- Using the Central GUI may be challenging because you can't modify LAG member speed individually.
- You can use the CLI too, but note that after release 2.5.9, Central will overwrite changes made in support mode, so you will have to complete the whole task in less time than it takes Central to identify the changes and change them back.
- It is also important to have a loop avoidance mechanism in place. I have MSTP (basic settings as provided by the Central GUI).
Remember that speeds cannot be mixed in a LAG - all members must be the same speed, so you cannot add a 25Gb member to a 10Gb LAG.
Upgrade from 10Gb to 25Gb
- On agg switch, edit config for first port (1/1/27): shutdown; speed auto 25g
- On access switch, swap in 25Gb transceiver first port (1/1/27)
At this stage, the LAG will be active on one link only at 10Gb still
- On agg switch, edit config to re-enable first port (1/1/27): no shutdown
& edit config for second port (1/1/28): shutdown; speed auto 25g; remove from lag
There will be some disruption here as the LAG changes from 10 to 25Gb; I usually see about 18 x 1s pings dropped.
LAG is now up with single 25Gb member
- On access switch, swap in 25Gb transceiver to second port (1/1/28)
- On agg switch, edit config for second port (1/1/28): no shutdown; lag 1
Now there is a LAG with 2 x 25Gb members.
Next step is 25 to 50Gb
The process is essentially the same again.
- On agg switch, edit config for first port (1/1/27): shutdown; remove speed setting (default is 50Gb)
- On access switch, swap in 50Gb transceiver first port (1/1/27)
- On agg switch, edit config to re-enable first port (1/1/27): no shutdown
& edit config for second port (1/1/28): shutdown; remove speed; remove from lag
There will be some disruption here as the LAG changes from 25 to 50Gb; I usually see about 18 x 1s pings dropped.
LAG is now up with single 50Gb member
- On access switch, swap in 50Gb transceiver to second port (1/1/28)
- On agg switch, edit config for second port (1/1/28): no shutdown; lag 1
Now there is a LAG with 2 x 50Gb members.
Q&A
Q: What speeds does the tri-speed R0M48A transceiver support?
A: 10, 25 or 50Gb
Q: Will the tri-speed R0M48A run in any port?
A: It will run only in SFP56 ports. These are found in the 6300, 6400, 8360 series of CX switches. IRIS can help you configure the correct transceivers for switches, along with the Transceiver Guide.
The output below is from a CX 6200 switch - it only has 10Gb SFP+ ports (not SFP56 like the 6300), and the 50Gb transceiver is specifically flagged as incompatible.
6200F-12# sh int 1/1/15 transceiver
------------------------------------------------------------------------------
Port Type Product Serial Part
Number Number Number
------------------------------------------------------------------------------
1/1/15 50G-SR R0M48A MYxxxxxxxx 1990-4803
6200F-12# sh int 1/1/15 transceiver detail
Transceiver in 1/1/15
Interface Name : 1/1/15
Type : 50G-SR / 50G SFP56 SR
Connector Type : LC
Transceiver Status : An HPE pluggable module that is incompatible with this interface
Wavelength : 850nm
Transfer Distance : 0.00km (SMF), 0m (OM1), 0m (OM2), 70m (OM3)
Diagnostic Support : None
Product Number : R0M48A
Serial Number : MYxxxxxxxx
Part Number : 1990-4803
Q: Is the tri-speed R0M48A auto-sensing?
A: No, it defaults to 50Gb, so you must specify 10 or 25Gb.
References
https://asp.arubanetworks.com/downloads;search=transceiver;sort=RELEASEDATE_DESC;fileTypes=DOCUMENT;products=Aruba%20Switches
AOS-Switch and AOS-CX Transceiver Guide
https://community.arubanetworks.com/viewdocument/aos-cx-1012-software-release-ente
AOS-CX 10.12 Training Materials
------------------------------
Richard Litchfield
Airheads MVP 2020, 2021, 2022
------------------------------