Floating Shaping in a nutshell?

Harvy66

Single WAN single LAN - Assigning Queues

Last floating wins, assuming no "quick"?
Interface specific rules that assign queues absolutely win?

Side question. I made 3 floating rules for TCP, UDP, and ICMP at the top and assigned them to separate queues. Since these are at the top of the floating rules, they should be over-ruled by any later rules. The reason I thought of this is because TCP, UDP, and ICMP tend to have different characteristics, and I wanted to place "unclassified" traffic into separate queues based on the general characteristic uses of the protocols. Sound logical?

bangheaduntildone

See: https://doc.pfsense.org/index.php/What_are_Floating_Rules

In particular:
_ Apply filtering in a "last match wins" way rather than "first match wins" (quick)_

So 1) Yes. 2) Yes if quick, otherwise other rules will still apply. So a (non-quick) floating "pass/allow" will still get over-ruled by a subsequent deny/block/reject if any such rule applies.

Side Qs: I'll let others speak for absolute latency issues, but have you considered just prioritizing the queues themselves? So VOIP queue gets top billing, etc etc. Trying to do it in two places gives you more chances to trick yourself later.

Harvy66

I do have separate queues for stuff like VoIP

I have queues for: games, dns, voip, ntp, p2p, web, vpn

Then I have catch-all rules for categorized TCP, UDP, and ICMP traffic, each in their own queues. ICMP is expected to be low bandwidth but latency sensative, so it gets a small portion of realtime. TCP and UDP are split 40/40 and P2P takes the remaining 20% of their parent queue.

pHigh LS: 19% <– Fixed bandwidth but latency sensitive. Shouldn't ever actually need link share, but just in case, it has some.
--qAck LS: 1% RT: 20%
--qGames LS: 1% RT: 10%
--qDNS LS: 1% RT: 2%
--qICMP LS: 1% RT: 1%
--qVoIP LS: 1% RT: 10%
--qNTP LS: 1% RT: 1%
pNormal LS: 50%
--qWeb LS: 50% CODEL
--qVPN LS: 50% CODEL
pLow LS: 31%
--qDefault LS: 40% CODEL
--qP2P LS: 20% CODEL
--qLowUDP LS: 40% RT: 5% CODEL<-- Tends to be stuff like random-port Mumble/TeamSpeak/etc servers. Stuff I can't know or I don't feel like polluting my rules with.

djvolt

Hi Harvy66

Could you backup your settings for this and put link?

Thank you :)

Derelict

Rule processing order:

Floating Rules
Interface Group Rules (Including OpenVPN, PPTP, L2TP, and PPPoE Tabs)
Interface Rules

Harvy66

–qLowUDP LS: 40% ~~RT: 5%~~

I later learned not to mix real time and link share. I seem to be still learning a some.

djvolt

This settings are for LAn or WAN interface?

I have this message:
03-02-15 01:00:39 [ There were error(s) loading the rules: pfctl: the sum of the child bandwidth higher than parent pHigh - The line in question reads [0]: ]

pHigh.JPG_thumb

Nullity

@djvolt:

This settings are for LAn or WAN interface?

I have this message:
03-02-15 01:00:39 [ There were error(s) loading the rules: pfctl: the sum of the child bandwidth higher than parent pHigh - The line in question reads [0]: ]

I assume that the leaf queues of qHigh may add up to over 100% or that you may have forgotten/mistyped some parameter(s).

Just to be safe, set the "Bandwidth" and link-share's m2 to the same in all queues.

Harvy66

While m2 link and linkshare are the same, I think the UI still wants linkshare to be set. Give it a try, just set it to the same.