Here's the scenario - AP is an ePMP 3000 running 4.4. Stations are a mix of Force 300 and Force 200, all running 4.4.
This week I swapped two of the legacy f200 units for f300 units. Identical configuration on old and new units. The two subscribers in question rely on MS Remote Desktop and since the swap, both report frequent dropping of RDP sessions. I have verified this behavior on-site with one of them. One of the subscribers is an individual running a single session, and the other is an office running 10~12 RDP sessions simultaneously.
Both CPE are in NAT mode, just as they were before when they were force 200 units. I'm blindly fidling with things now, as nothing has changed for these subscribers other than the hardware. Is there something that has changed in how the hardware handles nat session tracking or something?
I'll likely need to swap both back to f200 units over the weekend if I can't find any resolution.
If you're using NAT and DHCP DNS proxy or statically assigned DHCP DNS, this is broken and will be resolved in 4.4.1. This issue only effects AC clients... legacy clients do not have this issue.
For the time being, you'll need to statically assign your DNS servers on the clients downstream router or computer.
This doesn't sound like the issue to me. Name resolution works fine, it's that the nat tables in the radio seem to either fill up or flush at some interval, killing the RDP session. I'm waiting to hear back if it happens at a regular interval or randomly.
Unfortunately that would require a site visit to accomplish that can't happen until Monday. Odd that none of the other f300 subs I have on the same sector have any issues with the same config. Seems to just be RDP (and maybe some live-streaming app).
A followup on this - The only working solution is to bridge the CPE. This is not a DNS issue, but appears to be a NAT session tracking issue as best I can tell. Unfortunately with the CLI on ePMP there is no way to look at NAT sessions or tables.
Once the CPE is bridged, the issue goes away entirely.
I'm running into this same issue. It appears that every time the DHCP lease is renewed the interface bounces and the firewall rules are reloaded. I see this in the SM's log every time the issue happens:
Feb 25 19:43:06 hostname netifd: mgmt (2755): udhcpc: lease of 10.100.1.145 obtained, lease time 3600
Feb 25 19:43:06 hostname netifd: Interface 'mgmt' has lost the connection
Feb 25 19:43:06 hostname netifd: Interface 'mgmt6' is now down
Feb 25 19:43:06 hostname netifd: Interface 'mgmt6' is disabled
Feb 25 19:43:06 hostname dnsmasq[2467]: no servers found in /tmp/resolv.conf.auto, will retry
Feb 25 19:43:06 hostname miniupnpd[5121]: should send external iface address change notification(s)
Feb 25 19:43:06 hostname miniupnpd[5121]: ioctl(s, SIOCGIFADDR, ...): Cannot assign requested address
Feb 25 19:43:06 hostname miniupnpd[5121]: Failed to get IP for interface ath0.800
Feb 25 19:43:06 hostname miniupnpd[5121]: SendNATPMPPublicAddressChangeNotification: cannot get public IP address, stopping
Feb 25 19:43:06 hostname miniupnpd[5121]: PCPSendUnsolicitedAnnounce() IPv6 sendto(): Bad file descriptor
Feb 25 19:43:06 hostname miniupnpd[5121]: should send external iface address change notification(s)
Feb 25 19:43:06 hostname miniupnpd[5121]: PCPSendUnsolicitedAnnounce() IPv6 sendto(): Bad file descriptor
Feb 25 19:43:06 hostname netifd: Interface 'mgmt6' is enabled
Feb 25 19:43:06 hostname netifd: Interface 'mgmt6' is setting up now
Feb 25 19:43:06 hostname netifd: Interface 'mgmt6' is now up
Feb 25 19:43:06 hostname netifd: Interface 'mgmt' is now up
Feb 25 19:43:06 hostname dnsmasq[2467]: reading /tmp/resolv.conf.auto
Feb 25 19:43:06 hostname dnsmasq[2467]: using local addresses only for domain lan
Feb 25 19:43:06 hostname dnsmasq[2467]: using nameserver 8.8.8.8#53
Feb 25 19:43:06 hostname dnsmasq[2467]: using nameserver 8.8.4.4#53
Feb 25 19:43:06 hostname netifd: mgmt (2755): /sbin/uci: Parse error
Feb 25 19:43:06 hostname netifd: mgmt (2755): /sbin/uci: Parse error
Feb 25 19:43:11 hostname dnsmasq-dhcp[2467]: BOOTP(eth0) 5c:ac:4c:4d:cb:c2 no address configured
Feb 25 19:43:14 hostname firewall: Reloading firewall due to ifup of mgmt (ath0.800)