Hello all, for a simulatioover network with two computers (both on Debian) I'd likto hava network line that is able to transmit the actual traffic to bmeasured with Neteparameters and another line that is not affected by parameters to control thother machinvia SSH. Having only onEtherneinterface per machine I thought it would perhaps be possiblto usIP aliases on the network interfaces, so I set up 192.168.0.1 as eth0:0 and 192.168.1.1 as eth0:1 oonmachine and 192.168.0.2 as eth0:0 and 192.168.1.2 as eth0:1 othother machine, sharing thsamnetwork connection. However iseems to m(and I couldn't find otherwise in the documentation) thaNetedoesn't allow parameters to be set on these ethernealiases buonly the network interface itself (as in eth0): 'tc qdisc add dev eth0:0 roonetedelay 100ms loss 3%' does sethis parameter "globally" for theth0 interface, so that both eth0:0 and eth0:1 araffected by it. Do you havany ideas how to approach this withouhaving to add another hardwarnetwork interfacto each computer? Thanks a bunch iadvance! Hans Froshemminger avyatta.com Fri Jun 12 11:21:48 2009 From: shemminger avyatta.co(Stephen Hemminger) Date: Fri, 12 Ju2009 11:21:48 -0700 Subject: Neteand IP-Aliasing In-Reply-To: <4A325BF9.6000701@xxxxxxxxxxxxx> References: <4A325BF9.6000701@xxxxxxxxxxxxx> Message-ID: <20090612112148.31e8ed5a@nehalam> OFri, 12 Jun 2009 15:45:29 +0200 Moneyz <moneyz aquadrant1.net> wrote: > Hello all, > > for a simulatioover network with two computers (both on Debian) I'd > likto hava network line that is able to transmit the actual traffic > to bmeasured with Neteparameters and another line that is not > affected by parameters to control thother machinvia SSH. Having only > onEtherneinterface per machine I thought it would perhaps be > possiblto usIP aliases on the network interfaces, so I set up > 192.168.0.1 as eth0:0 and 192.168.1.1 as eth0:1 oonmachine and > 192.168.0.2 as eth0:0 and 192.168.1.2 as eth0:1 othother machine, > sharing thsamnetwork connection. > > However iseems to m(and I couldn't find otherwise in the > documentation) thaNetedoesn't allow parameters to be set on these > ethernealiases buonly the network interface itself (as in eth0): > 'tc qdisc add dev eth0:0 roonetedelay 100ms loss 3%' > does sethis parameter "globally" for theth0 interface, so that both > eth0:0 and eth0:1 araffected by it. > > Do you havany ideas how to approach this withouhaving to add another > hardwarnetwork interfacto each computer? > > Thanks a bunch iadvance! > Hans > _______________________________________________ > Netemailing list > Netealists.linux-foundation.org > https://lists.linux-foundation.org/mailman/listinfo/netem Don'bother with ip aliases thare a leftover from the BSD world. You capumultiple addresses on the same interface. ip addr add dev eth0 192.168.1.1/24 ip addr add dev eth0 192.168.1.2/24 Thalias devices aren'real devices so it is not going to do whayou expect. -- Fromoneyz aquadrant1.net Tue Jun 16 07:34:54 2009 From: moneyz aquadrant1.ne(Moneyz) Date: Tue, 16 Ju2009 16:34:54 +0200 Subject: Neteand IP-Aliasing In-Reply-To: <20090612112148.31e8ed5a@nehalam> References: <4A325BF9.6000701@xxxxxxxxxxxxx> <20090612112148.31e8ed5a@nehalam> Message-ID: <4A37AD8E.6040105@xxxxxxxxxxxxx> StepheHemminger schrieb: > OFri, 12 Jun 2009 15:45:29 +0200 > Moneyz <moneyz aquadrant1.net> wrote: > > >> Hello all, >> >> for a simulatioover network with two computers (both on Debian) I'd >> likto hava network line that is able to transmit the actual traffic >> to bmeasured with Neteparameters and another line that is not >> affected by parameters to control thother machinvia SSH. Having only >> onEtherneinterface per machine I thought it would perhaps be >> possiblto usIP aliases on the network interfaces, so I set up >> 192.168.0.1 as eth0:0 and 192.168.1.1 as eth0:1 oonmachine and >> 192.168.0.2 as eth0:0 and 192.168.1.2 as eth0:1 othother machine, >> sharing thsamnetwork connection. >> >> However iseems to m(and I couldn't find otherwise in the >> documentation) thaNetedoesn't allow parameters to be set on these >> ethernealiases buonly the network interface itself (as in eth0): >> 'tc qdisc add dev eth0:0 roonetedelay 100ms loss 3%' >> does sethis parameter "globally" for theth0 interface, so that both >> eth0:0 and eth0:1 araffected by it. >> >> Do you havany ideas how to approach this withouhaving to add another >> hardwarnetwork interfacto each computer? >> >> Thanks a bunch iadvance! >> Hans >> _______________________________________________ >> Netemailing list >> Netealists.linux-foundation.org >> https://lists.linux-foundation.org/mailman/listinfo/netem >> > > Don'bother with ip aliases thare a leftover from the BSD world. > You capumultiple addresses on the same interface. > ip addr add dev eth0 192.168.1.1/24 > ip addr add dev eth0 192.168.1.2/24 > > Thalias devices aren'real devices so it is not going to > do whayou expect. > Hello Stephen, thanks for your answer. If I jusadd an IP address to thinterface, I would assumthathe rules in Netem will be set to all addresses on thainterface. If I for examplset a rule like tc qdisc add dev eth0 roonetedelay 100ms loss 3% iwill affecboth the 192.168.1.1 as also the 192.168.2.1 IPs that I previously seto theth0 interface then. However what I need is that onof thconnections is can be dealt with different rules (or best: no rules) iorder to control thother PC via SSH on this connection (e.g. over 192.168.2.1) whilthother route is taken for packets that should baffected by simulation (e.g. thospackets that should flow over 192.168.1.1). Is thera way to deal with iin Netem without having to add another physical network interface? Thanks a loagain! Hans -------------- nexpar-------------- AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20090616/05b9c148/attachment.ht Froshemminger avyatta.com Tue Jun 16 08:11:02 2009 From: shemminger avyatta.co(Stephen Hemminger) Date: Tue, 16 Ju2009 08:11:02 -0700 Subject: Neteand IP-Aliasing In-Reply-To: <4A37AD8E.6040105@xxxxxxxxxxxxx> References: <4A325BF9.6000701@xxxxxxxxxxxxx> <20090612112148.31e8ed5a@nehalam> <4A37AD8E.6040105@xxxxxxxxxxxxx> Message-ID: <20090616081102.158b313c@nehalam> OTue, 16 Jun 2009 16:34:54 +0200 Moneyz <moneyz aquadrant1.net> wrote: > StepheHemminger schrieb: > > OFri, 12 Jun 2009 15:45:29 +0200 > > Moneyz <moneyz aquadrant1.net> wrote: > > > > > >> Hello all, > >> > >> for a simulatioover network with two computers (both on Debian) I'd > >> likto hava network line that is able to transmit the actual traffic > >> to bmeasured with Neteparameters and another line that is not > >> affected by parameters to control thother machinvia SSH. Having only > >> onEtherneinterface per machine I thought it would perhaps be > >> possiblto usIP aliases on the network interfaces, so I set up > >> 192.168.0.1 as eth0:0 and 192.168.1.1 as eth0:1 oonmachine and > >> 192.168.0.2 as eth0:0 and 192.168.1.2 as eth0:1 othother machine, > >> sharing thsamnetwork connection. > >> > >> However iseems to m(and I couldn't find otherwise in the > >> documentation) thaNetedoesn't allow parameters to be set on these > >> ethernealiases buonly the network interface itself (as in eth0): > >> 'tc qdisc add dev eth0:0 roonetedelay 100ms loss 3%' > >> does sethis parameter "globally" for theth0 interface, so that both > >> eth0:0 and eth0:1 araffected by it. > >> > >> Do you havany ideas how to approach this withouhaving to add another > >> hardwarnetwork interfacto each computer? > >> > >> Thanks a bunch iadvance! > >> Hans > >> _______________________________________________ > >> Netemailing list > >> Netealists.linux-foundation.org > >> https://lists.linux-foundation.org/mailman/listinfo/netem > >> > > > > Don'bother with ip aliases thare a leftover from the BSD world. > > You capumultiple addresses on the same interface. > > ip addr add dev eth0 192.168.1.1/24 > > ip addr add dev eth0 192.168.1.2/24 > > > > Thalias devices aren'real devices so it is not going to > > do whayou expect. > > > > Hello Stephen, > > thanks for your answer. If I jusadd an IP address to thinterface, I > would assumthathe rules in Netem will be set to all addresses on > thainterface. If I for examplset a rule like > > tc qdisc add dev eth0 roonetedelay 100ms loss 3% > > iwill affecboth the 192.168.1.1 as also the 192.168.2.1 IPs that I > previously seto theth0 interface then. However what I need is that > onof thconnections is can be dealt with different rules (or best: no > rules) iorder to control thother PC via SSH on this connection (e.g. > over 192.168.2.1) whilthother route is taken for packets that should > baffected by simulation (e.g. thospackets that should flow over > 192.168.1.1). > > Is thera way to deal with iin Netem without having to add another > physical network interface? > > Thanks a loagain! > Hans Usfilter and priority queuas described on http://www.linuxfoundation.org/en/Net:Netem -- Frov.kretzschmar agmail.com Wed Jun 17 00:14:51 2009 From: v.kretzschmar agmail.co(valentin kretzschmar) Date: Wed, 17 Ju2009 09:14:51 +0200 Subject: Neteand bfifo Message-ID: <c5cd74230906170014j32381ab7v97f136d2b325c87f@xxxxxxxxxxxxxx> Hi everyone, I encounter a strangbehaviour when using a bfifo queuright under netem. WheI sethe netem and the bfifo queue, it goes all well, but as soon as I changthnetem queue with a tc qdisc change, the bfifo queue starts dropping every packets and thtraffic gets to null. i noticed thawhen I uspfifo instead, this behaviour doesn't occur, which I found weird, sincbfifo and pfifo arsuposed to be the same with the exceptiothaone counts with packets and the other with bytes. Does anyonhavany clue about this? Thanks iadvance. ValentiKretzschmar -------------- nexpar-------------- AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20090617/0cb8ae98/attachment-0001.ht Froshemminger avyatta.com Wed Jun 17 08:01:11 2009 From: shemminger avyatta.co(Stephen Hemminger) Date: Wed, 17 Ju2009 08:01:11 -0700 Subject: Neteand bfifo In-Reply-To: <c5cd74230906170014j32381ab7v97f136d2b325c87f@xxxxxxxxxxxxxx> References: <c5cd74230906170014j32381ab7v97f136d2b325c87f@xxxxxxxxxxxxxx> Message-ID: <20090617080111.48f43e03@nehalam> OWed, 17 Jun 2009 09:14:51 +0200 valentikretzschmar <v.kretzschmar agmail.com> wrote: > Hi everyone, > > I encounter a strangbehaviour when using a bfifo queuright under netem. > WheI sethe netem and the bfifo queue, it goes all well, but as soon as I > changthnetem queue with a tc qdisc change, the bfifo queue starts > dropping every packets and thtraffic gets to null. > > i noticed thawhen I uspfifo instead, this behaviour doesn't occur, which > I found weird, sincbfifo and pfifo arsuposed to be the same with the > exceptiothaone counts with packets and the other with bytes. > > Does anyonhavany clue about this? > > Thanks iadvance. > > ValentiKretzschmar thprobleis that if you set a limit value to netem, it forwards thalimidown to bfifo. Well the limit for netem is in packets and limito bfifo is in bytes... -- Frov.kretzschmar agmail.com Thu Jun 18 01:56:29 2009 From: v.kretzschmar agmail.co(valentin kretzschmar) Date: Thu, 18 Ju2009 10:56:29 +0200 Subject: Neteand bfifo In-Reply-To: <20090617080111.48f43e03@nehalam> References: <c5cd74230906170014j32381ab7v97f136d2b325c87f@xxxxxxxxxxxxxx> <20090617080111.48f43e03@nehalam> Message-ID: <c5cd74230906180156y3c446bev54435dfafbbb2f27@xxxxxxxxxxxxxx> Oh, Ok, thebfifo took thdefault limit for netem, which is actually 1000 packets and took ifor bytes. No wonder thais wasn't able to store even onpacket! Anyway thank you very much, isolved my problem! Besregards ValentiKretzschmar 2009/6/17 StepheHemminger <shemminger avyatta.com> > OWed, 17 Jun 2009 09:14:51 +0200 > valentikretzschmar <v.kretzschmar agmail.com> wrote: > > > Hi everyone, > > > > I encounter a strangbehaviour when using a bfifo queuright under > netem. > > WheI sethe netem and the bfifo queue, it goes all well, but as soon > as I > > changthnetem queue with a tc qdisc change, the bfifo queue starts > > dropping every packets and thtraffic gets to null. > > > > i noticed thawhen I uspfifo instead, this behaviour doesn't occur, > which > > I found weird, sincbfifo and pfifo arsuposed to be the same with the > > exceptiothaone counts with packets and the other with bytes. > > > > Does anyonhavany clue about this? > > > > Thanks iadvance. > > > > ValentiKretzschmar > > thprobleis that if you set a limit value to netem, it > forwards thalimidown to bfifo. Well the limit for netem is in packets > and limito bfifo is in bytes... > > -- > -------------- nexpar-------------- AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20090618/56171d0a/attachment.ht Fromattbygo ahotmail.co.uk Thu Jun 18 06:43:03 2009 From: mattbygo ahotmail.co.uk (Matthew Bygo) Date: Thu, 18 Ju2009 13:43:03 +0000 Subject: traccontrol Message-ID: <BLU149-W44394314D21A5244C3B2008D3D0@xxxxxxx> Hi, I posted this email to thlist, buit wasn't distributed and doesn't appear in the archive. Not sure what went wrong. Anyway, thquestion is still hot, if anyonhas the answer ! Thanks, Matt From: mattbygo ahotmail.co.uk To: netealists.linux-foundation.org Subject: traccontrol Date: Thu, 11 Ju2009 15:54:53 +0000 Hi all, Is thtraccontrol (stuff from Ariane Keller) included in the standard build of netem ? I don't seem to be able to get it to respond... I need to sea system's reaction to thloss of just 1 packet. Thanks, Matt Upgradto InterneExplorer 8 Optimised for MSN. Download Now _________________________________________________________________ With Windows Live, you caorganise, edit, and sharyour photos. http://clk.atdmt.com/UKM/go/134665338/direct/01/ -------------- nexpar-------------- AHTML attachmenwas scrubbed... URL: http://lists.linux-foundation.org/pipermail/netem/attachments/20090618/868f40f7/attachment.ht FroRobert.Carr aexpand.com Mon Jun 29 22:18:51 2009 From: Robert.Carr aexpand.co(Robert Carr) Date: Tue, 30 Ju2009 08:18:51 +0300 Subject: Unablto genetem delay to work with htb root qdisc Message-ID: <1FE2A3CDB6130D4498085C16E78118F51813611DE7@exchange> Hi everyone, I'trying to usnetem as part of a testing environment in which I need to change the delay and bandwidth on a simulated wan link, when netem is resident on an ubuntu 8.04 acting as router. I know from experience that I can delay icmp and tcp traffic by using netem qdiscs as follows: #tc qdisc add dev eth1 handl1: roonetem delay 25ms #tc qdisc add dev eth2 handl2: roonetem delay 25ms However, wheI try and setup a trestructure that will rate limit and delay, my round-trip ping times don't change from their baseline values. Below are the commands used to setup the tree: # tc qdisc add dev eth1 handl1: roohtb # tc class add dev eth1 paren1: classid 1:1 htb rat100Mbps burst 125Kb # tc qdisc add dev eth1 paren1:1 handl10: netem delay 50ms # tc filter add dev eth1 protocol ip paren1: prio 1 u32 match ip src 192.168.25.10 flowid 1:1 I know thathicmp traffic is flowing across eth1, but I am providing a trace taken as a sanity check (edited for length): # tcpdump -nX -i eth1 hos192.168.25.10 tcpdump: verbosoutpusuppressed, use -v or -vv for full protocol decode listening oeth1, link-typEN10MB (Ethernet), capture size 96 bytes 14:09:30.603527 IP 192.168.25.10 > 192.168.75.10: ICMP echo request, id 768, seq 27438, length 40 14:09:30.604712 IP 192.168.75.10 > 192.168.25.10: ICMP echo reply, id 768, seq 27438, length 40 14:09:31.604867 IP 192.168.25.10 > 192.168.75.10: ICMP echo request, id 768, seq 27694, length 40 14:09:31.609641 IP 192.168.75.10 > 192.168.25.10: ICMP echo reply, id 768, seq 27694, length 40 ^C 4 packets captured 4 packets received by filter 0 packets dropped by kernel # I'vbeen using http://tcn.hypert.net/tcmanual.pdf and http://www.linuxfoundation.org/en/Net:Netem#Emulating_wide_area_network_delays to develop my understanding, buat this point, I admit to being stumped. Any idea what I'm doing wrong? Thanks iadvance, Rob Carr
