Impact of changing 'Default MTU Size' from 9578 to 9416

[its_Tricky83]

So I've just noticed the big red writing under 'Switch Settings' which is informing us that our two aggregated core switches don't support our configured MTU Size.

Can anybody tell me if there is any adverse impact in changing our Default MTU Size from 9578 down to 9416? 

I'm wondering if I can simply change this to 9416, hit apply and it not have any noticeable disruption on our network.

[cmr]

@its_Tricky83 if the 425s are the core of the network, then they are already limiting the maximum packet size.  Therefore changing the max size down to the supported level should not have any effect on traffic between VLANs.  It could however impact traffic within a VLAN if there are devices configured to the maximum size.

i.e. Server with MTU 9500 and iSCSI SAN with MTU 9500 will currently talk fine if the traffic does not go through the 425s.  If you changed it for all switches to 9416 then the iSCSI packets would get fragmented.  If, however either device has an MTU of 9416 or lower then that traffic would be unaffected by the change.

It isn't common to see large MTUs other than for iSCSI or sometimes video streaming.  I'd check any of those devices that are on your network and not already passing the traffic through the 425s before making the change.

[PhilipDAth]

Personally - I would change the MTU to the supported value.  Change any servers/SANS/etc to using a 9,000 byte MTU (a more standard size for end devices).

Sizes larger than 9000 are often only used by network devices so they can do encapsulation and add extra things to the packet (like double-tagged frames).

[its_Tricky83]

Thank you both @cmr and @PhilipDAth, 

We have a relatively simple topology.
The 425s are at the core of this site and stacked and providing Layer 3 switching, which are then each connected to separate CPE MikroTik (VRRP Warm Spare) routers with static routes over 1/1Gbps NCS connections connecting back to the core of the network.

Then we have about 30 x 225 edge switches LACP'd back to the 425 core stack. 
Edge wise, just about all devices and system are workstations with the odd IP camera and UPS powering the switches. 

All our servers and core systems are upstream at datacentres.

So I'm somewhat convinced it should be fine to drop the MTU down. I don't even recall setting the MTU to 9578 in the first place. Maybe the MTU size came over from the Catalyst migration to Meraki but there are no more Catalysts in play in this site.

I might drop it down to the max compat size of 9416, given that there is a fair amount of routing going on and therefore might have some encapsulation/extra-stuff going on. 

[Justin_Blignaut]

I would also drop it down to the supported MTU. From past experience I can confirm that traffic across these devices will be impacted, when you click save. In my case it was around 15min before all services recovered.