Slow fetch in pfSense shell + slow squid downloads
-
I noticed this issue while adding a squid proxy to our pfSense box. It looks like the pfSense download speeds are far slower than what hosts behind the pfSense on the LAN are getting.
If I fetch a file to /dev/null on the pfSense box, it goes at about 800 kBps. At the same level of WAN saturation the same file downloads at 2.5MB/s over wireless from my computer. When I enable the squid proxy all downloads seem to go the speed of the fetch on the pfSense box (which I guess makes sense since it's proxying the file through the pfSense box).
What could this be?
-
What could this be?
Slow disk?
SSH in and run:
squidclient -h <lan_ip> -p 3128 mgr:info</lan_ip>and check the Median Service Times for any outliers.
-
when using /dev/null , it shouldn't matter if the disk is slow or fast
-
Some more discoveries from the past couple hours that may help me or others find the issue:
It definitely has nothing to do with squid - everything weird I experience is also experienced by the pfSense locally with fetch.
It has a very negative effect on a download where two machines have high RTT. For an S3 test with us-west bucket (we are in SF) it is able to get up to 9.5MB/s (even though from a machine on the network I can get 13MB/s). For a us-east bucket it sticks right around 800kBps.
Is there something different the pfSense does to handle it's own traffic / squid's traffic? I should mention we have two gateways and I use two gateways groups (prefer one, failover to the other) splitting the DHCP space in half to assign some users to each.
-
I was going to setup a squid proxy on another box on the network, but I'd really like to have squid on the same box as the pfSense. This is really the only issue standing in my way. Are there other things I can do to see why the pfSense box would have a slow speed locally?
