Posted on April 22, 2019 4:53 pm
 |  Asked by Michael Soto
Print Friendly, PDF & Email

This might be a dumb question: Why is the MLAG reload-delay by default so long? What are the risks of setting it shorter?

I was surprised to see that the delay on the 1G fixed-port DCS-7020TR-48-R was 10 minutes. And the modular 7504-N is 30 minutes. That’s a lot of additional time to sit and wait after-hours during a maintenance window..

Posted by Neha Goyal
Answered on April 22, 2019 5:16 pm

Hi Michael,

Reload-delay is the time period during which Mlag agent’s SysdbPlugin will keep the ports in err-disabled state and is important for graceful reformation of MLAG peer where we can sync all the P2p states. Modular switches take longer to come up due to the number of linecards, ports, agents etc that need to be initialized and hence in order to give the modular switch enough time to stabilize from a hardware perspective the recommendation is to configure a higher timeout than for fixed configuration switches. The risks of setting it shorter is that agents and hardware might not be ready to forward traffic and hence you might experience traffic loss.

Starting from EOS-4.15.2 release, default reload-delay is automatically adjusted to account for platform ( modular – 1800s in sand platform, 1200s in sequoia and non-modulars – 300s ).

Hope this helps.


Answered on April 23, 2019 5:34 am

Hi Michael.

The reload delay is to prevent the switch from receiving traffic before all the interfaces are up and protocols have converged (otherwise, the traffic can be potentially blockholed if the network has not fully converged). This is particularly important on modular switches, hence the longer period for the modular switches. The default is typically the recommended setting, but in case the default timer is to long for your particular deployment, you may adjust this by changing the configuration:

Post your Answer

You must be logged in to post an answer.