CPU Utilisation

[Wurzi]

Hi all,

We have a Meraki MX250 acting as a Hub in our SD-WAN deployment, and we're seeing CPU usage consistently hovering around 75%. We're still onboarding sites, so I'm concerned this may worsen.

Here's what I've observed:

• WAN throughput isn't very high: approx. 2 x 250Mb load balanced across 2 x 1Gbps connections.

• "Windows File Sharing" is the top application listed under traffic analytics.

• The routing table exceeds 2000 prefixes.

I am looking to reduce the number of prefixes and considering add the IP of the Windows server to Trusted IP Addresses/Subnets.

Any other thoughts?

Many Thanks in advance. 

[jimmyt234]

What MX firmware are you running? There are some improvements in 19.1.9 if you are not already up to that version yet, example being:

• Corrected an issue that resulted in MX75, MX85, MX95, MX105, MX250, and MX450 appliances reporting erroneously higher device utilization on the Meraki Dashboard.

[Wurzi]

Running Current version: MX 18.211.5.1, I am being told by the MX the firmware is up to date

[RWelch]

If you were to go to Organization > Monitor > Firmware upgrades you will see the Latest Firmware Versions available and you can see the versions for each listed: 

Stable
Stable release candidate
Beta
Other available versions

MX18.211.6 is the stable version.
MX19.1.9 is the stable release candidate.

@jimmyt234 mentioned the improvements in MX19.1.9 and I would agree with his suggestion if you wanted to give MX19.1.9 a try.  You can roll back the firmware up to 14 days if you aren't pleased with the improvements.

 

 

[RWelch]

And you can always SUBSCRIBE to the Firmware Upgrades Feed in the Meraki Community if you want firmware release announcements sent to your e-mail inbox - it's a handy way to see when firmware releases are posted/published (released).

[DarrenOC]

How was the MX250 sized during pre-sales?

How many sites are you deploying in total?

[Wurzi]

We have 178 branches. Currently, we have 168 connected.

[DarrenOC]

How many WAN connections does each Branch site have?

[DarrenOC]

Hi @Wurzi ,  worth reviewing the below link to check you’re not hitting the Max number of supported tunnels for your Hub MX

https://docs.google.com/spreadsheets/d/11eeTkL_xALtnGCNwPZRyJ6kJtCrVyzdetnUsDus6HrY/htmlview

[DarrenOC]

Which, for the MX250 is recommended to be 1000

[Wurzi]

Thanks, checked. It looks to be the maximum unique sessions at the moment working with TAC

[GIdenJoe]

Darren made an excellent point with the amount of tunnels since if both are running in NAT mode and each side has 2 ISP's you can take your site count and quadruple that number to know your tunnel count.

However maximum sessions is indeed another issue to tackle with.  It must be that you have too many hosts each making small concurrent sessions.  In that case if you want to keep the SD-WAN design you may need to change up to one-armed concentrators instead and group branches each to their own preferred concentrator.

[RaphaelL]

Open a ticket to support and ask them to review detailed stats. They will tell you the bottleneck ( CPU , Mem , PPS , tunnels and so on )

[Steve_PF]

Another thing to consider is the high volume of small packets .  We ran to an issue where CPU of our hubs went to the roof after an Citrix upgrade.  Since we coudn't do aything about the trafic patern changes, we ended by doubling our MX250 to split the loads.