[SOLVED] Traffic graph weirdness

maverick_slo

done
https://redmine.pfsense.org/issues/7751

maverick_slo

Hi!

Any idea if this is pfsense bug or upstream?
I would really like this to be resolved in 2.4…

PiBa

Maybe you could install "nload" on ssh/console?

pkg add http://pkg.freebsd.org/freebsd:11:x86:64/latest/All/nload-0.7.4_1.txz

Start it use arrows to switch to desired nic/vlan and compare with traffic graphs.

It should pull statistics from sysctl just like pfSense.. And if also wrong then points a little to freebsd.
Though i suppose it could still be a issue introduced by pfSense's changes somewhere..

maverick_slo

Hehe :)
So here we go, guess pfsense bug after all….

Untitled.png_thumb

PiBa

Oké i guess its still not 100% definitive…

Pfsense calls sysctl with:

CTL_NET, PF_LINK, NETLINK_GENERIC, IFMIB_IFDATA, {ifidx}, IFDATA_GENERAL

Where it retrieves the data: bytes, packets, collisions, multicasts, mtu
https://github.com/pfsense/FreeBSD-ports/blob/devel/devel/php56-pfSense-module/files/pfSense.c#L3210

While nload calls sysctl with:

CTL_NET, PF_ROUTE, 0, 0, NET_RT_IFLIST, 0

Where it retrieves the data: bytes, packets, errors
https://github.com/liuyang1/nload/blob/master/src/devreader-bsd.cpp#L108

So its similar but not the same, and that might be for a good reason. I'm not exactly sure.. Though i guess most of those extra data items are not needed for traffic graphs..

maverick_slo

Jarred is on the case
Maybe just fewer flags and thats it.

jdillard

I think this is the issue jahonix was referring to: https://redmine.pfsense.org/issues/3314

If that's the case someone else might pick up the ticket, but for now I'll try and find the commit with the patch.

edit: I tried running:

git log --since "MAY 18 2015" --until "MAY 20 2015" --pretty=format:"%Cred%h%Creset -%C(yellow)%d%Creset %s %Cgreen(%ad) %C(bold blue)<%an>%Creset"

on the following repos:


pfsense/pfsense
pfsense/FreeBSD-ports
pfsense/FreeBSD-src

without any luck finding the patch so far. I may have accidentally missed it or be taking the wrong approach.

PiBa

Found it: https://github.com/pfsense/pfsense-tools/commit/ad55e66a3fb817f8e54bae7bf38306039b82e3ad

Or: https://github.com/pfsense/FreeBSD-src/commit/939f11b196e7d22a1ce3da8533e97ff3e92510b6#diff-437f2b0c100f75faf75f7c5b60eca6a4R1156

PiBa

2.3:

 			ifp->if_opackets++;
 			//ifp->if_omcasts += mcast;
 			//ifp->if_obytes += len;

2.4:

			if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1);
			if_inc_counter(ifp, IFCOUNTER_OBYTES, len);
			if_inc_counter(ifp, IFCOUNTER_OMCASTS, mcast);

In 2.4 it looks different.. probably needs some // as well.. https://github.com/pfsense/FreeBSD-src/blob/RELENG_2_4/sys/net/if_vlan.c#L1132

jahonix

I love this kind of collaboration, really great!

jdillard

Nice work PiBa! And thanks for the lead jahonix!

I updated the ticket with the new info: https://redmine.pfsense.org/issues/7751

maverick_slo

Thanks guys!
This us why I prefer pfsense to other fw's… ;)

jahonix

@jared: I didn't do a thing despite remembering that I had this problem in the past myself.
And someone gave me a smite for it anyways. ;-)

I'd love to contribute more actively but my programming abilities are elsewhere (Crestron mainly).

maverick_slo

Resolved with todays snap!

Thanks guys!

Raul Ramos

Hi

My Lan is a parent interface of 2 Vlan's. When someone transfere something from internet on one of the Vlan's interface the Lan interface have the same bandwidth usage (if no one use Lan otherwise it sum the bandwidth). Ideally, in my case, should be the bandwidth used in Vlan not sum in the Parent interface.

Some stuff i can point is the rule from Vlan to internet have a limiter.

The double bandwidth is solved in Vlan's interfaces.