Packet loss on PPPoE interface
-
Hello,
I'm a total noob when it comes to traffic shaping. Running Dual-WAN, single LAN through the wizard setup. We actually are running a Dual-WAN. One is PPPoE, the other is DHCP. In Pftop it's showing a lot of dropped packets on the primary PPPoE WAN connection. I'm routing all https traffic through that connection, as it's the bigger pipe and our finance team got tired of their https connections to the bank dropping out when it switched over to the secondary WAN. Any suggestions as to why I'm getting so many dropped packets or where I should look? I'm really grateful for any ideas.
pfTop: Up Queue 1-28/28, View: queue
QUEUE BW SCH PR PKTS BYTES DROP_P DROP_B QLEN BORR SUSP P/S B/S
root_pppoe1 970K hfsc 0 0 0 0 0 0
qInternet 970K hfsc 0 0 0 0 0
qACK 174K hfsc 8308 428K 2 94 0
qDefault 388K hfsc 17318 21M 28309 36120K 19
qP2P 38800 hfsc 0 0 0 0 0
qOthersHigh 87300 hfsc 2258 1140K 32 21369 10
qOthersLow 97000 hfsc 0 0 0 0 0
root_em4 679K hfsc 0 0 0 0 0 0
qInternet 679K hfsc 0 0 0 0 0
qACK 122K hfsc 483 28146 0 0 0
qDefault 61110 hfsc 863 108K 0 0 0
qP2P 27160 hfsc 0 0 0 0 0
qOthersHigh 61110 hfsc 615 179K 0 0 0
qOthersLow 67900 hfsc 0 0 0 0 0
root_em3 1000M hfsc 0 0 0 0 0 0
qLink 200M hfsc 14136 1573K 0 0 0
qInternet 10M hfsc 0 0 0 0 0
qACK 1966K hfsc 1994 118K 0 0 0
qP2P 437K hfsc 0 0 0 0 0
qOthersHigh 983K hfsc 16081 20M 0 0 0
qOthersLow 1092K hfsc 0 0 0 0 0
root_em2_vlan4 1000M cbq 0 810 675K 0 0 0 0 0
qLink 200M cbq 2 30 4684 0 0 0 0 0
qInternet 10M cbq 0 0 0 0 0 0 0
qACK 1966K cbq 6 190 12004 0 0 0 0 0
qP2P 437K cbq 0 0 0 0 0 0 0
qOthersHigh 983K cbq 4 590 658K 0 0 0 242 0
qOthersLow 1092K cbq 3 0 0 0 0 0 0 0 -
I could be very wrong, but dropped packets on a queue are completely normal, even necessary when said queue becomes full. I don't see any packets dropped on the interface, just the child queues (though, I dunno if that is relevant).
I would pay more attention to the interface (ifconfig) packet error stats.
-
The default queue size is only 50 packets. That may be too small for your bandwidth. You could also try enabling CoDel, then you don't need to worry about queue size.
-
Thanks for the replies. I got very excited when I read up on Codel. It seems like the easiest and most effective option. But, it didn't work as expected. I enabled Codel on my primary WAN connection (the PPPoE one) and my speed test improved a bit, but my ping times doubled during the speed test. I switched back to HFSC and ran another speed test. My speeds were slower, but my ping times stayed stable. I've tried fiddling with enabling Codel on all of the interfaces (WAN, WAN2 (non-PPPoE) and LAN) and I keep getting the same results. Ping times go up when I run a speed test, but when it change it back to HFSC the ping times stay the same. I also tried adding rate limits in the interface config as some have suggested, but it didn't help either. Any suggestions as to why Codel doesn't perform as well as HFSC?
-
Thanks for the replies. I got very excited when I read up on Codel. It seems like the easiest and most effective option. But, it didn't work as expected. I enabled Codel on my primary WAN connection (the PPPoE one) and my speed test improved a bit, but my ping times doubled during the speed test. I switched back to HFSC and ran another speed test. My speeds were slower, but my ping times stayed stable. I've tried fiddling with enabling Codel on all of the interfaces (WAN, WAN2 (non-PPPoE) and LAN) and I keep getting the same results. Ping times go up when I run a speed test, but when it change it back to HFSC the ping times stay the same. I also tried adding rate limits in the interface config as some have suggested, but it didn't help either. Any suggestions as to why Codel doesn't perform as well as HFSC?
Probably because of a misconfiguration or confusion about what it is capable of. Codel itself only controls queue depth. It really shouldn't affect bitrate.
CoDel does not help latency caused by download saturation. Only rate-limiting can help there.
When you talk about latency, was this measured during the upload or download part of the speedtest.Were you using stand-alone codel (codelq) or the "codel active queue" toggle?
-
Thanks, Nullity. I was doing a download speed test when I was noticing the latency. I was also only using stand-alone codel, not the codel active queue. What do you recommend? I'm still obviously learning about this stuff.
-
Thanks, Nullity. I was doing a download speed test when I was noticing the latency. I was also only using stand-alone codel, not the codel active queue. What do you recommend? I'm still obviously learning about this stuff.
If you have complex needs, like running multiple cloud backups while wanting to watch Netflix and play games, then use HFSC/CBQ/PRIQ/FAIRQ and "codel active queue".
Otherwise "codelq" might work well.
Read http://www.linksysinfo.org/index.php?threads/qos-tutorial.68795/ for a great intro into traffic-shaping/QoS. He does a great job demistifying misconceptions and explains the differences between the solutions to fixing download & upload latency/bandwidth problems.