> It's not. > > # ethtool -k eth1 > Offload parameters for eth1: > rx-checksumming: on > tx-checksumming: on > scatter-gather: on > tcp-segmentation-offload: off > udp-fragmentation-offload: off > generic-segmentation-offload: off > generic-receive-offload: off > large-receive-offload: off > ntuple-filters: off > receive-hashing: off
And what's on eth0 and eth2? What I don't understand is your interfaces file, where you have isp1 on eth0 and isp2 on eth1, while in another place you have isp1 -> eth1 and isp2 -> eth2. Is this all correct? Simon > > -Olivier > ----- Mail original ----- > De: "Tom Eastep" <[email protected]> > Ã: "Shorewall Users" <[email protected]> > Envoyé: Mercredi 4 Décembre 2013 00:56:39 > Objet: Re: [Shorewall-users] Multiple ISP + traffic shapping = > poor download speed > > On 12/3/2013 2:03 PM, [email protected] wrote: >> Hello, >> >> Thanks for the great Shorewall which has replaced my hard to maintain >> home-made scripts. >> >> First, what works. >> >> Our local network is 10.48.X.X with multiple vlan, each on a dedicated >> interface. We use Shorewall 4.4.11 from Debian Squeeze. >> >> We have a 2 ISP: >> - isp1 : an optical fiber provider with 10 Mbps. >> - isp2 : a DSL provider with 15Mbits/1Mbits. >> >> We use isp2 as the default outgoing provider. The isp1 provider is used >> for "critical" services (SSH...) and for incoming connections (VPN...). >> >> Our interfaces file : >> ======================== >> isp1 eth0 detect >> logmartians,nosmurfs,routefilter=0,tcpflags >> isp2 eth1 detect >> logmartians,nosmurfs,routefilter,tcpflags >> ======================== >> >> Here is our providers file: >> ======================== >> #NAME NUMBER MARK DUPLICATE INTERFACE GATEWAY >> OPTIONS COPY >> isp1 1 0x100 - eth1 37.X.X.X >> track,loose - >> isp2 2 0x200 - eth2 217.X.X.X >> track,balance - >> ======================== >> >> Here is an extract of our tcrules file: >> ======================== >> ###################################################################################################################### >> #MARK SOURCE DEST PROTO DEST SOURCE USER >> TEST LENGTH TOS CONNBYTES HELPER >> # PORT(S) PORT(S) >> >> # ISP1 DNS => ISP1 >> 256 0.0.0.0/0 37.X.X.X >> 256 $FW 37.X.X.X >> >> # ISP2 DNS => ISP2 >> 512 0.0.0.0/0 127.X.X.X >> 512 $FW 127.X.X.X >> >> # Google DNS => ISP1 >> 256 0.0.0.0/0 8.8.8.8,8.8.4.4 >> 256 $FW 8.8.8.8,8.8.4.4 >> >> # VPN IPsec (out) => ISP1 >> 256 0.0.0.0/0 0.0.0.0/0 udp 500,4500 >> 256 $FW 0.0.0.0/0 udp 500,4500 >> >> # Force one host to ISP1 >> 256 10.48.1.10 0.0.0.0/0 >> >> # Force all SSH to ISP1 >> 256 0.0.0.0/0 0.0.0.0/0 tcp 22 >> 256 $FW 0.0.0.0/0 tcp 22 >> ======================== >> >> Yesterday we added VoIP. To do so, we force traffic from our Asterisk >> server to go throw ISP1 with a dedicated public IP and force the traffic >> from this dedicated public IP to go to Asterisk server (with IP >> filtering for security). This works too. >> >> Now, my problem is to put QoS (using TC_ENABLED=Internal). I try many >> configuration but always have the same problem: once the isp1 interface >> is listed in tcdevices, we have poor download speed. Even with/without >> other TC configuration. >> >> Here is our tcdevices file: >> ======================== >> #NUMBER: IN-BANDWITH OUT-BANDWIDTH OPTIONS REDIRECTED >> #INTERFACE INTERFACES >> 1:isp1 10240kbit 10240kbit >> ======================== >> >> We use an external server to test download speed with IP 5.X.X.X so we >> added in tcrules: >> ======================== >> 256 0.0.0.0/0 5.X.X.X >> $FW 0.0.0.0/0 5.X.X.X >> ======================== >> >> The results are: >> - without isp1 in tcdevices => more than 1MB/s (bytes measured with wget >> command) >> - with isp1 in tcdevices => less than 300 kB/s >> >> If I change bandwidth of isp1 to something more than 70000kbit, all goes >> right... Other lower value have the same problem but with different >> download speed (seems proportional to the interface speed). >> >> Here is a result of the following command: tc -s -d class show dev isp1 >> ======================== class htb 1:1 root rate 10240Kbit ceil >> 10240Kbit burst 1598b/8 mpu 0b overhead 0b cburst 1598b/8 mpu 0b >> overhead 0b level 7 >> Sent 1111091 bytes 11680 pkt (dropped 0, overlimits 0 requeues 0) >> rate 83656bit 124pps backlog 0b 0p requeues 0 >> lended: 0 borrowed: 0 giants: 0 >> tokens: 17781 ctokens: 17781 >> ======================== >> >> Rates seems to be OK. >> >> Have someone the same problem? >> > > Sounds like Shorewall FAQ 97a. > > -Tom > -- > Tom Eastep \ When I die, I want to go like my Grandfather who > Shoreline, \ died peacefully in his sleep. Not screaming like > Washington, USA \ all of the passengers in his car > http://shorewall.net \________________________________________________ > > > ------------------------------------------------------------------------------ > Sponsored by Intel(R) XDK > Develop, test and display web and hybrid apps with a single code base. > Download it for free now! > http://pubads.g.doubleclick.net/gampad/clk?id=111408631&iu=/4140/ostg.clktrk > _______________________________________________ > Shorewall-users mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/shorewall-users > ------------------------------------------------------------------------------ Sponsored by Intel(R) XDK Develop, test and display web and hybrid apps with a single code base. Download it for free now! http://pubads.g.doubleclick.net/gampad/clk?id=111408631&iu=/4140/ostg.clktrk _______________________________________________ Shorewall-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/shorewall-users
