MX Cluster (Warm Spare) Upgrade

[renanmonteiro]

Good evening,

I have a MX84 cluster on my DC, as a VPN hub. It will be necessary to perform a forced upgrade of the two appliances and I would like to know if there is any "optimal" strategy to perform the upgrade of them.

I was thinking of undoing the Warm Spare, add the secondary MX in a temporary network, just for the upgrade, after updating it, it would be waiting to finish the update of the primary and then return the HA.

There is a possibility of having to reset the boxes, however, disregarding this, would this strategy be a good ideal?

Thanks

[AjitKumar · Accepted solution]

Hi @renanmonteiro 

I understand "Dashboard" will take care of this challenge with Zero downtime.

Could you please refer the following Url (If already not)

https://documentation.meraki.com/Architectures_and_Best_Practices/Cisco_Meraki_Best_Practice_Design/...

[cmr]

We have upgraded HA routed pairs dozens of times and for most clients there will be no noticeable downtime.  What we have seen on a busy site is:

• Cisco IP SIP phones registered to a server on a remote site may show 'registering' for 15 seconds or so.
• Citrix XenApp thin clients may see a 'reconnecting' pop up from the taskbar but applications continue to operate normally.

For PCs we've not seen any impact.  I also think that ongoing SIP phone calls aren't affected but I will need to test to be sure.

[swifty]

Is an MX HA pair stateful ??

Same thing here, users asking for impact statement.

According to the doc referenced elsewhere
https://documentation.meraki.com/Architectures_and_Best_Practices/Cisco_Meraki_Best_Practice_Design/...

"  the dashboard will automatically take steps to minimize downtime when upgrades are performed to ensure a zero-downtime MX upgrade "

Will a tcp state table be copied across or will tcp connections be reset ?

My understanding ref @cmr 's comment, SIP phones - once call setup has taken place - will be keeping the call going between endpoints and unless the call flow, or one of the endpoints, is via the device being upgraded there is no impact.
If the SIP server is via the device in q, I can see a hit, but I would have thought it would be very quick if 'zero downtime' is the reality.

[cmr]

@swifty the only flow that we have seen interrupted is RTSP camera feeds where the client has to refresh the view after the failover.

[swifty]

Hi All

I have just come off the phone to Meraki TAC.

The failover is not stateful.
So TCP resets all round, but device available achieved by failover and fallback.

Ian

[swifty]

@cmr Thanks again. 👍
All went fine, customer reports no users reported any disruption to service.
Not that I think they would have noticed anyway.

[renanmonteiro]

Thanks, 

We did the upgrade yesterday and worked beautifully, meraki's support checked the upgrade had been completed and we finished activity very quickly.