PFSense 1.2.2 - 30MB Fiber Link - Uploads are being limited to between 3-10MB.

Perry

I would first test with a live boot cd to see if it is a hardware problem. There is always Commercial Support if it's business critical.

jmcvay

Cide,

Did you ever resolve your issue?

I am having a similar issue with a 10Mbps Fiber Link from XO Communications, Downstream is fine, put our upstream caps at around 3Mbps.

Hooking a laptop directly the the ethernet handoff works correctly, so I'm assuming it iss something within PFSense 1.2.3 Configuration or possibly hardware related.

cmb

Check the interfaces under Status > Interfaces. Look for errors, collisions where there shouldn't be.

jmcvay

No collisions reported.

WAN:

Media 100baseTX <full-duplex>In/out packets 11696702/9921518 (1.68 GB/3.56 GB)
In/out errors 0/0
Collisions 0

LAN:

Media 100baseTX <full-duplex>In/out packets 12370614/13220069 (3.88 GB/2.90 GB)
In/out errors 0/0
Collisions 0

XO and I set our routers to 100FDX since we were having the same issue with them set to auto-negotiate.</full-duplex></full-duplex>

danswartz

I'd be curious to see what happened if you set both ends to 100/half rather than 100/full. Also, can you get interface stats from their side, not just yours?

cmb

It still sounds like a duplex mismatch. It sounds like you have a switch between your WAN and the Cisco router, if that's the case you must force speed and duplex on the switch port, not the router (and if you do on the router, you must on its switch port as well - if you're going to force, make absolutely sure everything is forced, or you will end up with a duplex mismatch and serious performance problems). If you have an unmanaged switch where you can't force the port, you must run autonegotiate on everything.

I've seen some VIA NICs that refuse to force speed and duplex when it's manually set. In those cases, the interface status showed incorrectly though so that isn't the same case here. I'd switch out the NICs with a different chipset and see what happens, if your switch ports and router are definitely all set correctly.

danswartz

Yeah, I agree.

jmcvay

Some more oddities.

We sent out an email to a very large amount of individuals on Friday's and I noticed my bandwidth meter's transmit going through the roof.

It was actually capping around 13Mbps (Our fiber is supposed to cap at 10Mbps)

I cannot replicate this on speedtest.net or speakeasy.net.

Another oddity is that we have 4 interfaces. WAN, LAN, DMZ, and Wireless.

WAN and LAN are pretty self explanatory.

I have the DMZ setup for Webserver with a public IP, and restrict the DMZ from accessing the LAN.

The Wireless interface connects directly to a LAN port on an old Westell Wireless HUB.

The DMZ and Wireless interfaces use the same make/model NIC.

I connected my laptop the the WLAN and ran a speedtest, and low and behold I am hitting close to our 10Mbps.

Some details:

ISP Router - Set for 100FDx

Cross Over cable directly connected between ISP Router and WAN Interface.

WAN
bge0: flags=8843 <up,broadcast,running,simplex,multicast>metric 0 mtu 1500
options=9b <rxcsum,txcsum,vlan_mtu,vlan_hwtagging,vlan_hwcsum>media: Ethernet 100baseTX <full-duplex>status: active

LAN
bge1: flags=8843 <up,broadcast,running,simplex,multicast>metric 0 mtu 1500
options=9b <rxcsum,txcsum,vlan_mtu,vlan_hwtagging,vlan_hwcsum>ether 00:22:19:d5:ee:93
inet 192.168.168.1 netmask 0xffffff00 broadcast 192.168.168.255
inet6 fe80::222:19ff:fed5:ee93%bge1 prefixlen 64 scopeid 0x4
media: Ethernet autoselect (1000baseTX <full-duplex>)
status: active

DMZ
em0: flags=8843 <up,broadcast,running,simplex,multicast>metric 0 mtu 1500
options=19b <rxcsum,txcsum,vlan_mtu,vlan_hwtagging,vlan_hwcsum,tso4>ether 00:1b:21:39:55:c1
inet 10.0.0.1 netmask 0xffffff00 broadcast 10.0.0.255
inet6 fe80::21b:21ff:fe39:55c1%em0 prefixlen 64 scopeid 0x1
media: Ethernet autoselect (1000baseTX <full-duplex>)
status: active

WIRELESS
em1: flags=8843 <up,broadcast,running,simplex,multicast>metric 0 mtu 1500
options=19b <rxcsum,txcsum,vlan_mtu,vlan_hwtagging,vlan_hwcsum,tso4>ether 00:1b:21:39:57:73
inet6 fe80::21b:21ff:fe39:5773%em1 prefixlen 64 scopeid 0x2
inet 172.16.0.1 netmask 0xffffff00 broadcast 172.16.0.255
media: Ethernet autoselect (100baseTX <full-duplex>)
status: active

LAN is connected to a Managed HP ProCurve Switch 2810-24G with Duplex settings set to Auto.</full-duplex></rxcsum,txcsum,vlan_mtu,vlan_hwtagging,vlan_hwcsum,tso4></up,broadcast,running,simplex,multicast></full-duplex></rxcsum,txcsum,vlan_mtu,vlan_hwtagging,vlan_hwcsum,tso4></up,broadcast,running,simplex,multicast></full-duplex></rxcsum,txcsum,vlan_mtu,vlan_hwtagging,vlan_hwcsum></up,broadcast,running,simplex,multicast></full-duplex></rxcsum,txcsum,vlan_mtu,vlan_hwtagging,vlan_hwcsum></up,broadcast,running,simplex,multicast>

jmcvay

Another update, this appears to only be affecting HTTP Upstream.

SMTP and FTP are able to consistently hit our upstream cap and maintain it.

What would be causing only HTTP to perform subpar? We have tested with QoS disabled, and QoS enabled with everything as default and HTTP set to higher priority, no difference.

danswartz

Hmmm, I assume this is bulk SMTP and/or FTP traffic? What kind of traffic is the HTTP? Is it bulk too? Or lots of smaller packets? Maybe a limitation on the number of packets/sec you can push thru?

jmcvay

The email I described was likely 2MB in size, but was sent to about 800 recipients.

The FTP and HTTP test consisted of uploading the same 1GB file to a hosting company I use. FTP thru FileZilla and HTTP through cPanel.

FTP hits cap, but HTTP still struggles around 1.5-3Mbps just as illustrated on Speedtest.net and Speakeasy.net.

danswartz

Is it possible to try a different http client? It would eliminate questions about, say, send window size or whatever…

jmcvay

I've run the speedtest.net, speakeasy.net, and a cPanel upload in both IE and Firefox. :-\

danswartz

Oh, sorry, was going by the comment about the http test being via cpanel.

danswartz

I guess without seeing a wireshark capture, it's really hard to suggest anything more…

Efonnes

Are you using any kind of HTTP proxy? (like squid, for example)

jmcvay

I had installed squid at one point in time, but removed it.

I'm almost wondering if one of my packages didn't uninstall properly. I currently do not have any packages installed. I will be performing a fresh install this coming Friday when I have a maintenance window. If I exhibit the same symptoms I may give the BETA a try.