MX - Odd correlation between WAN latency and Cellular signal

[RaphaelL]

Hi ,

I have over 1600 spokes ( MX68CW ) running on their WAN1-WAN2 + Cellular backup ( integrated SIM ). 

However this site is pretty odd. The WAN latency seems to be a perfect copy of the Cellular signal stats... ( the site is not running on cellular ) 

I don't have any issues at all , I just find that super odd...  

Running 18.107.6. Has anyone ever seen that ?

Case already open to do a spot check.

[PhilipDAth]

I wonder if your WAN backhaul provider uses some kind of RF technology in their network or the same cellular backhaul.

[RaphaelL]

So when the cellular connection went down, the latency and loss went to 0  😃

Mar 2 15:42:47

Cellular

Cellular connection down

model: Integrated, provider: XXXX, connection: lte

Confirmed that the LTE and WAN1 have nothing in common. When we experienced the latency and loss we would notice on the dashboard the cellular bouncing multiple time per minute between 'connecting' to 'ready'. Everytime it would flip , we had packet loss. 

I can't repro that at all... very strange.

[PhilipDAth]

That must be a bug.

[RaphaelL]

Just found another site with ugly cellular signal that matches the WAN latency... 

[RaphaelL]

I was able to run a ping to one the SVI on that MX. The moment the cellular state changed , I experienced the latency : 

[RaphaelL]

Upgrading to MX 18.107.9 ( latest patch ) doesn't fix this issue.  Support is pointing to our "large" L3 firewall rule base ,which currently has 70-75 rules. We didn't have these issues before upgrade to MX18. 

I will keep digging. 

[PhilipDAth]

Really?  They are saying it can't cope with 70 to 75 firewall rules?  It doesn't sound right to me.

[RaphaelL]

From what I understood , Support says they had a similar case with a customer with a "large" rulebase and everytime the cellular interface flaps , the MX has to "recalculate" the rulebase, and that would explain why they are seeing CPU spikes or peg @100% during the same timeslot.