@StephenNorbury 

I'm chasing the same thing on the ER605V1, I have a pretty detailed thread going here https://community.tp-link.com/en/business/forum/topic/600644 

I've had it looked at a couple of times by support, a couple of beta test loads, but the usage just keeps climbing linearly over time.

If you just want to prevent this happening again until a fix is found, just set a weekly reboot at 3AM.  This will reset things before starvation occurs.  I lived like that for over a year, I've undone it to help try to find the root cause...and I'm crashing about every 30 days too (it used to be every 21 days).

Hello  @StephenNorbury,

You may upgrade to this Beta firmware to have a try:

Solution to Repeating Notification of "[Gateway] was connected" Issue on ER605 V2

If you still get the notifications after upgrades, please feel free to let us know.

Hello @StephenNorbury,

Sorry, my fault. Please allow me to confirm some details regarding the router disconnected issue you mentioned:

1. When the router was disconnected, would all of your network devices drop the network connection? Or would only the router get disconnected from the controller?

2. What did you have to do to recover the connection when the issue happened? Would it work automatically after some time?

Did you have any Gateway ACL and mDNS settings?

Please try to provide more details (setup and network diagram if possible) as possible as you can to help us better understand the issue you encountered.

  @StephenNorbury 

What does the status (memory/cpu) graph look like immediately after a reboot?  Having your memory at 90% is a red flag for me.  My bet is after a reboot, that drops to just under 50% and then creeps up until it stops working again.

This is a live plot from my ER605v1..I expect it to start failing again in 10-14 more days

  @GoPokes1 

I personally believe the high CPU is an indirect symptom of contrained memory, so the unit is continually swapping out parts of memory to disk in order to try to continue...and as the memory leak continues...eventually the process crashes which results in a the router no longer responding.   In most cases, it's the Omada app that dies first (which is why it shows as disconnected, but generally keeps forwarding packets), but eventually I believe I have seen the DHCP process crash as well.

  @GoPokes1 

I am working on identifying the culprit with support.

Until then I suggest you implement a weekly reboot schedule at 3am one day.  This kept me going for a year until I had the SNMP history to back up my mem leak theory....now I am running crash to crash for the greater good.

GoPokes1 wrote

I think you've describe the behaviors accurately including the eventual consequence of DHCP failing. 

Is there a firmware (beta or otherwise) to address the high CPU / memory leak issues?