MX Global Bandwidth Limit issue

[PatR-31C]

Ran into an issue last week that hit us pretty hard and left us scratching our heads.   MX100 was reconfigured to use a single VLAN over an access port instead of the multiple VLAN trunks that is was using previously.   The day after the change, users were complaining about slow internet speeds.   Speed tests show 1Mbps download on a 300Mbps circuit.   Of course, panic ensued and rather than bore everyone with the gory details, in the end it turned out that the per-client global bandwidth limit setting (20Mbps per client + SpeedBurst) was the culprit.   Strange thing is that this particular MX was our first Meraki device and has had that setting for 3 years now.    Other MXs at sites in our organization have the same setting.    What could possibly cause that setting on that MX to behave differently?     Any thoughts?

[DensyoV]

Hi,

Even if you have 20Mbps per client + SpeedBurst setting, it should not cause to limit the clients' throughput to only 1 Mbps.

You can check the changelogs if there any other settings have been changed. ie. on the traffic shaping page, the client tracking option on the MX, and on the port settings of the switch connecting to the MX. 

If you face the issue again or if you are want to test it (putting the 20Mbps per client + SpeedBurst back), I recommend calling the Meraki support to help you with the checking as this is definitely not expected behavior.

thanks,

[PhilipDAth]

>MX100 was reconfigured to use a single VLAN over an access port instead of the multiple VLAN trunks that is was using previously

The limit is imposed on the MAC address.  Any chance you introduced a layer 3 routing device, so all the user traffic now appeared to be coming from a single MAC address?

[PatR-31C]

Thank you for the response.

The device the MX was plugged into prior to the change and after the change was an HP 8200zl switch.   It is providing L3 routing.    During the confusion (there was and still is a lot) we swapped out the HP 8200 switch with a MS425 and had the same issue.

I understand the point you are making regarding the limit being placed on the MAC address.   The issue I am having trouble resolving in my brain is that we have MXs at other sites with the same bandwidth limit setting that are plugged into similar HP L3 switches that don't exhibit the same behavior.   Some use trunks, some use access ports.   None of them have ever had this issue.

When we put in the case with Meraki support the tech was quick to point out the bandwidth limit which got us back on track quickly, but it took us a while to get to that point because the HP switch was the focus of our troubleshooting efforts based on the fact the other MXs weren't having any issues  iPerf testing clearly showed the effect of the bandwidth limit, but left me really scratching my head.