Is 100GB QSFP Hot-Swappable Manual Handling Guidelines Recommendation

[octavianUK]

After working extensively with Cisco equipment, I recently had a debate with a close friend about the handling of 100GB QSFP transceivers. I argued that these transceivers are hot-swappable, meaning the port doesn’t need to be "powered down" for removal. To check the QSFP model, you can simply pull the tab latch to disconnect the transceiver, inspect its make and model, and then reinsert it. Everything should resume working within a couple of minutes.

However, my friend referenced the Cisco Optical Transceiver Handling Guide, which suggests a more cautious approach. According to the guide:

"Before extracting a transceiver from the host platform, the host should verify that the port is no longer carrying active traffic and is ready for extraction."

At the same time, the guide also states:

"The QSFP-DD, QSFP, and SFP transceiver modules are hot-swappable and connect the electrical circuitry of the system with an optical external network."

These two statements seem to contradict each other. My friend’s concern is the potential impact on the lifespan of the QSFP transceiver if the first guideline is not followed.

Can someone clarify this? Is there any real impact on the QSFP’s lifetime if the port isn't powered down before swapping the module?

[Joseph W. Doherty]

I don't see the contradiction.

As a QSPF is probably an infrastructure port, just yanking the transceiver, or shutting the port, if there's active traffic is likely to be somewhat disruptive to the network.

[octavianUK]

My understanding of hot-swappable is that you don't need to switch off the system or the port if you need to disconnect it. Of course there is disruptive to the network, I'm not disputing that. the question is, is there any impact on the QSFP lifetime ?

Regarding the contradiction, if it's hot-swappable then you shouldn't turn off the port in order to swap/change the QSFP. If you need to turn off the port, then the term hot-swappable cannot be used. 

[Joseph W. Doherty]

I don't disagree but to put it another way the two statements don't contradict each other as the first only mentions active traffic, nothing about electrical considerations.  Even if you're thinking, doesn't active traffic impact electrical signals, sure it does, but consider rarely is a port not passing some traffic, like CDP/LLDP even when not carrying active traffic.

So, I take the first statement as a recommendation not to yank a transceiver if it's carrying active traffic, basically the same recommendation applies to shutting down the port, just a good recommendation.

The second statement explicitly notes the transceiver is hot swappable.  Which for just physical considerations, allows you to physically safely ignore the above recommendation concerning active traffic.

So, I don't see violation of the recommendation to consider active traffic would physically impact the transceiver.  The first recommendation doesn't explicitly say insure the port is physically shut down first.  Of course, granted a shut port would verify lack of active traffic, but so would doing a couple of show interfaces and observing neither ingress or egress counters are incrementing.

[octavianUK]

I see. The way I saw the recommendations, it was mainly due to the temperature the QSFP's can get, it is recommended to power down the port, but is not a must. And I believe there is no negative impact on the lifetime of the QSFP if it's unplugged while is live (do you think I'm right or wrong?). 

[Leo Laohoo]

Are Cisco optics hot-swappable (aka OIR)?  

Yes and no.  It depends on the platform, particularly the operating system and the firmware version. 

Yes, they are meant to be OIR.  But in reality there are some firmware version which negates OIR because of bad coding.  In those cases, inserting the optic alone will blow out the processes that monitors the port and causes the port not to provide power to the optic.  With those bugs, a complete cold reboot (pull out the power supplies) is the best known "solution".  

[octavianUK]

so what would you suggest in this scenario. 

I will need to check around 500 QSFP's if the right ones were used. 

What is the best way of doing that?

[Leo Laohoo]

---

@octavianUK wrote:
I will need to check around 500 QSFP's if the right ones were used. 

---

The command "sh inventory" is the first command to use.  

Any optic(s) not in the list, visually inspect it because if the optic is not getting "recognized" by the switch then there is a strong chance the bug has been hit.