A long post, needed to give some explanation 
:
All routers run v6.18 with latest firmware.
All routers have interface queue set to ‘hardware-only’
ALL wireless routers have wireless fp package enabled.
All bridges have rtsp protocol enabled and 'Allow Fast Path" is enabled everywhere. No further settings.
One router “A” (rb1100AH);
Ether9: MTU=1500, L2 MTU=1589
vlan2000 : MTU=1500, L2 MTU=1594, ID=2000, Use Service Tag=yes Interface=Ether9
vlan300-trunk : MTU=1500, L2 MTU=1594, ID=300, Use Service Tag=yes Interface=Ether9
vlan301-trunk : MTU=1500, L2 MTU=1590, ID=301, Use Service Tag=yes Interface=vlan300-trunk
vlan310 : MTU=1500, L2 MTU=1590, ID=310, Use Service Tag=yes Interface=vlan300-trunk
vlan320 : MTU=1500, L2 MTU=1586, ID=320, Use Service Tag=yes Interface=vlan301-trunk
vlan325 : MTU=1500, L2 MTU=1586, ID=325, Use Service Tag=yes Interface=vlan301-trunk
vlan326 : MTU=1500, L2 MTU=1586, ID=326, Use Service Tag=yes Interface=vlan301-trunk
vlan327 : MTU=1500, L2 MTU=1586, ID=327, Use Service Tag=yes Interface=vlan301-trunk
vlan330 : MTU=1500, L2 MTU=1586, ID=330, Use Service Tag=yes Interface=vlan301-trunk
vlan340 : MTU=1500, L2 MTU=1586, ID=340, Use Service Tag=yes Interface=vlan301-trunk
Each vlan that is not a trunk-vlan has a dhcp server running.
Resuming;
Ether9 carries 2 vlans; vlan2000 & vlan300-trunk
vlan300 carries 2 vlans; vlan 301 & vlan 310
vlan301-trunk carries 6 vlans; vlan320, 325, 326, 327, 330 & 340
Ether9 of rb1100AH connects by utp to Ether1 of “B”
Router “B” (rb911);
Ether1 and wlan1 of rb911 are both in bridge. (So all traffic passes transparent)
On all 3 interfaces (wlan1, ether1 and the bridge) settings are; MTU=1500 and L2 MTU=1600
Wireless; Bridge mode, duo chain, 5Ghz-onlyN, NV2 with encryption enabled, 20/40Mhz eC and some rates set; 180Mbps in both direction stable conn. rate.
wlan1 connecto to 9km remote router “C”
Router “C” (rb911);
Same config as “B” but in mode ‘station-bridge’
Ether1 connects with utp to Ether1 of router “D”
Both router “B” and “C” are ‘transparant bridges’ thus no vlan settings required.
Router “D” (rb800);
Ether1: MTU=1500, L2 MTU=1600, connect to ether1 of router “C”
Ether2: MTU=1500, L2 MTU=1600, connect to ether1 of router “E”
Ether3: MTU=1500, L2 MTU=1600, connect to ether2 of router “F”
wlan1 & wlan2 are both configured AP-bridge to serve clients; MTU=1500, L2 MTU=1600
Ether1, ether2 & ether3 are all in one bridge “BH-bridge”; bridge MTU=1500, L2 MTU=1600
wlan1 & wlan2 are both in one bridge “AP-bridge”; bridge MTU=1500, L2 MTU=1600
vlan300-trunk [(interface=BH-bridge); MTU=1500, L2 MTU=1596.] is configured here. (The allocation of this vlan300 trunk creates the option to ‘tap’ the embedded vlan310 to make it available in the “AP-bridge”)
vlan310 [(interface=vlan300-trunk); MTU=1500, L2 MTU=1596.] is placed in the “AP-bridge” to ensure connected CPE’s of AP radio’s can request and be delivere CPE (=client ID) IP address.
vlan301 with the other embedded (encapsulated?) vlans travel through the “BH-bridge” to become available in other routers. Outside the scope of this post.
Ether9 or “A” has IP xx.xx.54.1/29
Bridge of “B” has IP xx.xx.54.2/29
Bridge of “C” has IP xx.xx.54.3/29
Ether1 of “D”'s “BH-bridge” has IP xx.xx.54.4/29
I can ping and reach (winbox) all these routers by these IP’s. Ping time between A and B is in the order of 2-5ms
On vlan310 dhcp-server IP yy.yy.yy.2 is assigned to “AP-Bridge” of router “D”
I can now ping from router “A” to this bridge ip address yy.yy.yy.2 and get same results. 2-6ms
So far so good!
BUT, from any client CPE connected to one of the AP radio’s of router “D” I can ping this AP-bridge IP yy.yy.yy.2 but ping times are all over the place. Average from 10ms to regular 20-80ms and regurlarly 200ms or higher!
All clients have wireless fp package enabled, hardware queue on the wlan interface (which is not in a bridge, and dhcp-client) and MTU=1500 and L2 MTU=1600 (Ether1 side = LAN = MTU=1500 and L2 MTU=1598)
One AP radio serves about 16-18 clients, the other serves about 37-40 clients.
AP1 (16-18 clients) works with duo chain AP radio and antenna. Clients are a mix of duo and single chain radio CPE’s
AP2 (37-40 clients) works with duo chain AP radio on 2x vertical polarizid antenna and all CPOE’s are single chain radios.
Now; BEFORE implementing 6.18 and the wireless fp package and the hardware-queue on the interface this AP was considered to be one of my best performing AP’s. Always fast responce and short and stable ping times with good througputs, no traffic cuts.
Now, AFTER the implemantation of 6.18 and fp wherever possible the whole network goes crap. People can’t make no more voip calls, complain about slow and breaking internet and yes indeed, even the opening of winbox session to CPE’s is slow. Throughputs towards clients is slow, some MB only and very fluctuating, even althouht the total produced traffic of this AP is only some 20-30Mb’s, nothing for a rb800 and the main backhaul has no problems with it. Before I could easy push 80-100Mb over the link and I still can. But the client network went bananas!
What happened! My feeling goes towards the MTU settings, that now with the vlans and because the fastpath everywhere is not good anymore?
Before with the normal wireless package and the normal interface settings there was nothing wrong.
I’ll guess the combination of NV2 tdma period size, with the MTU setting and/or fastpath settings and the use of vlan as transport LAN creates the problem?
Anybody any ideas or help or suggestions?