Nexus LACP fast timer vs. ISSU

[Niels Grimm]

Hello,

I have question or maybe a "feature request" regarding LACP timing on Nexus switches.

One of the best features of Nexus 5k and 7k is ISSU for us but if there are endstations (server, firewalls etc.) connected via portchannel with LACP fast timing (short timeout = 1s) ISSU won't be hitless (no graceful as in L3) even the Nexus port is configured with long timeout (normal =30s). Unfortunately I found no configuration way to overule the fast timing of the endstation configuration. The Nexus always accept the fast timing from the opposite.

Do you know a way to overrule that?

If not is there a plan to do that in future releases?

Kind regards,

Niels

[MARK BAKER]

Niels,

Did you ever find a way to get your end device to use the long timeout? We have an EMC NAS server that is showing short timeout as well. We are researching how to configure this to use the long timeout. Even though we haven't found the answer yet, I have to assume it can be changed especially since it shows the timeout value in the LACP display for the link.

Thanks,
Mark

[l33d5un1t3d]

Neils/Mark,

did you ever get a fix...........I have the same issue which is preventing ISSU working, it's only the EMC NAS interfaces that we have this with, the EMC consultant was next to useless???

[Steve Fuller]

Hi,

I don't believe there's anything that can be done to resolve this issue as this is how the IEEE standard specifies it should operate. According to clause 5.4.1 of the IEEE 802.1AX standard:

5.4.1 LACP design elements

The following considerations were taken into account during the development of the protocol described in

this subclause:

[..]

d) Periodic transmission of LACPDUs occurs if the LACP_Activity control of either the Actor or the

Partner is Active LACP. These periodic transmissions will occur at either a slow or fast transmission

rate depending upon the expressed LACP_Timeout preference (Long Timeout or Short Timeout) of

the Partner System.

Essentially if Cisco choose to ignore the preference of the partner then their IEEE 802.1AX implementation will be non compliant.

I've seen the recommendations from Cisco that long timers should not be used when performing ISSU, but what is the actual issue you face? Do you get a specific error message telling you that ISSU is not supported or is it a warning and ISSU then continues?

If it's the fact that the Nexus won't perform the ISSU when the EMC is using LACP short timeout, is there the option to make a change to your processes when performing software upgrades on the Nexus switch? If you were to disable the interface to the EMC prior to the ISSU do you still see the same issue?

Regards

[l33d5un1t3d]

Hi Steve,

we aborted the ISSU after the warning not knowing whether it would impact all live services, and we had this as a non-disruptive change CR. In fact we had already succesfully performed ISSU already on our DR infrastrucure, but didn't have EMC in the equation. We were hpoing to get round this by having the EMC guys go to long timeouts, but the couldn't for some reason?

Change windows on this scale are few and far for us, and we don't have lab kit try this on, ISSU was a major selling point to get some expensive  hardware in, but we (or the VAR) didn't know of this particular "gotcha".........so will need to try and engineer a solution or take input from someone who's done it already.....thanks for the input to date.

[Steve Fuller]

Hi,

I completely understand about change control and change windows. Necessary, but still the bane of our lives.

I've not actually tried ISSU with devices attached running short timers, but assume the problem to be that as the Nexus stops sending LAPCDU, the EMC will fail the member links connected to the Nexus after approximately 3-seconds. Just talking purely hypothetically, I'd argue that the short timers have served their purpose and the traffic has been moved to the remaining operational links. This of course assumes that all links of the EMC are not connected to a single device.

As I mentioned already, the other option might be to down the links first. This will allow you to fail over in a coordinated way with the storage team, and will also fail over quicker as the EMC will fail immediately on link down rather than waiting for the LACPDU timeout.

If I get a chance at any point I'll have a mess around with this in the lab and see what happens.

Regards

[MARK BAKER]

We were able to correct the ISSU incompatibility by configuring the EMC storage unit to use long timers. I didn't configure the EMC. I asked the person who maintains it to do it if it could be done. He found the setting to change it and all is good now. I sent an email to our EMC admin asking if he remembers how he changed it. I'll give an update once I hear from him.

FYI... In case you are not aware, there are some ISSU compatibility check commands that you can run to verify the switch is ISSU ready before you attempt an upgrade. Two that come to mind are LACPD and spanning-tree.

Mark

[MARK BAKER]

The EMC unit is an VNX5500. Attached is the procedure used to change the LACP timeout to long.

Hope this helps,
Mark