On Wed, Jun 26, 2024 at 02:15:02PM GMT, Eelco Chaudron wrote: > On 5 Jun 2024, at 22:23, Adrian Moreno wrote: > > > Test simultaneous IPFIX and local sampling including slow-path. > > I guess Ilya's comments on this patch summarize most of my comments. I had > one small additional question. See below. > > //Eelco > > > Signed-off-by: Adrian Moreno <amore...@redhat.com> > > --- > > tests/system-common-macros.at | 4 ++ > > tests/system-traffic.at | 105 ++++++++++++++++++++++++++++++++++ > > 2 files changed, 109 insertions(+) > > > > diff --git a/tests/system-common-macros.at b/tests/system-common-macros.at > > index 2a68cd664..22b8885e4 100644 > > --- a/tests/system-common-macros.at > > +++ b/tests/system-common-macros.at > > @@ -378,3 +378,7 @@ m4_define([OVS_CHECK_GITHUB_ACTION], > > # OVS_CHECK_DROP_ACTION() > > m4_define([OVS_CHECK_DROP_ACTION], > > [AT_SKIP_IF([! grep -q "Datapath supports explicit drop action" > > ovs-vswitchd.log])]) > > + > > +# OVS_CHECK_EMIT_SAMPLE() > > +m4_define([OVS_CHECK_EMIT_SAMPLE], > > + [AT_SKIP_IF([! grep -q "Datapath supports emit_sample" > > ovs-vswitchd.log])]) > > diff --git a/tests/system-traffic.at b/tests/system-traffic.at > > index bd7647cbe..babc56b56 100644 > > --- a/tests/system-traffic.at > > +++ b/tests/system-traffic.at > > @@ -8977,3 +8977,108 @@ OVS_WAIT_UNTIL([cat p2.pcap | grep -E "0x0050: > > *0000 *0000 *5002 *2000 *b85e *00 > > > > OVS_TRAFFIC_VSWITCHD_STOP > > AT_CLEANUP > > + > > +AT_SETUP([emit_sample]) > > +OVS_TRAFFIC_VSWITCHD_START() > > +OVS_CHECK_EMIT_SAMPLE() > > + > > +ADD_NAMESPACES(at_ns0, at_ns1) > > +NS_CHECK_EXEC([at_ns0], [sysctl -w net.ipv6.conf.all.disable_ipv6=1], [0], > > [ignore]) > > +NS_CHECK_EXEC([at_ns1], [sysctl -w net.ipv6.conf.all.disable_ipv6=1], [0], > > [ignore]) > > + > > +ADD_VETH(p0, at_ns0, br0, "10.1.1.1/24", "e4:11:22:33:44:55") > > +ADD_VETH(p1, at_ns1, br0, "10.1.1.2/24", "e4:11:22:33:44:66") > > + > > +NS_CHECK_EXEC([at_ns0], [arp -s 10.1.1.2 e4:11:22:33:44:66]) > > +NS_CHECK_EXEC([at_ns1], [arp -s 10.1.1.1 e4:11:22:33:44:55]) > > + > > + > > +AT_CHECK([ovs-appctl vlog/set dpif:dbg dpif_netdev:dbg > > ofproto_dpif_upcall:dbg]) > > + > > +AT_CHECK([ovs-vsctl -- --id=@br0 get Bridge br0 \ > > + -- --id=@ipfix create IPFIX targets=\"127.0.0.1:4739\" > > \ > > + -- create Flow_Sample_Collector_Set id=1 bridge=@br0 > > ipfix=@ipfix, local_sample_group=10 \ > > + -- create Flow_Sample_Collector_Set id=2 bridge=@br0 > > ipfix=@ipfix, local_sample_group=12], > > + [0], [ignore]) > > + > > +AT_DATA([flows.txt], [dnl > > +in_port=ovs-p0,ip > > actions=sample(probability=65535,collector_set_id=1,obs_domain_id=1431655765,obs_point_id=1717986918),output(port=ovs-p1,max_len=100) > > +in_port=ovs-p1,ip > > actions=sample(probability=65535,collector_set_id=2,obs_domain_id=2290649224,obs_point_id=2576980377),output(port=ovs-p0,max_len=100) > > +]) > > + > > +AT_CHECK([ovs-ofctl add-flows br0 flows.txt]) > > + > > +OVS_DAEMONIZE([ovstest test-psample > psample.out], [psample1.pid]) > > + > > +NS_CHECK_EXEC([at_ns0], [ping -q -c 1 10.1.1.2 | FORMAT_PING], [0], [dnl > > +1 packets transmitted, 1 received, 0% packet loss, time 0ms > > +]) > > + > > +m4_define([SAMPLE1], [group_id=0xa > > obs_domain=0x55555555,obs_point=0x66666666 > > .*icmp.*nw_src=10.1.1.1,nw_dst=10.1.1.2]) > > +m4_define([SAMPLE2], [group_id=0xc > > obs_domain=0x88888888,obs_point=0x99999999 > > .*icmp.*nw_src=10.1.1.2,nw_dst=10.1.1.1]) > > +AT_CHECK([grep -E 'SAMPLE1' psample.out >/dev/null]) > > +AT_CHECK([grep -E 'SAMPLE2' psample.out >/dev/null]) > > + > > +AT_CHECK([ovs-ofctl dump-ipfix-flow br0 | sed 's/tx pkts=[[0-9]]*/tx > > pkts=24/' | sed 's/tx errs=[[0-9]]*/tx errs=0/'], [0], [dnl > > Why are we making tx pkts always 24, and are we waiving any potential tx > errors? > Is this something you have seen not being consistent, if so any idea why? >
Yes. By the time we request the statistics, IPFIX cache should contain the flows but they might not have been sent yet. When they are sent they will fail (because IPFIX target is localhost and it returns the socket returns an error if the listening socket does not exist) so the number of errors might vary. I can look deeper into making it more consistent but it looked like a quick workaround. > > +NXST_IPFIX_FLOW reply (xid=0x2): 2 ids > > + id 1: flows=1, current flows=0, sampled pkts=1, ipv4 ok=1, ipv6 ok=0, > > tx pkts=24 > > + pkts errs=0, ipv4 errs=0, ipv6 errs=0, tx errs=0 > > + id 2: flows=1, current flows=0, sampled pkts=1, ipv4 ok=1, ipv6 ok=0, > > tx pkts=24 > > + pkts errs=0, ipv4 errs=0, ipv6 errs=0, tx errs=0 > > +]) > > + > > +AT_CHECK([ovs-appctl lsample/show br0], [0], [dnl > > +local sample statistics for bridge "br0": > > +- Collector Set ID: 1 > > + Local Sample Group: 10 > > + Total number of bytes: 98 > > + Total number of packets: 1 > > + > > +- Collector Set ID: 2 > > + Local Sample Group: 12 > > + Total number of bytes: 98 > > + Total number of packets: 1 > > +]) > > + > > +# Disable trunc feature to force traffic to go through slow path. > > +AT_CHECK([ovs-appctl dpif/set-dp-features br0 trunc false]) > > + > > +AT_CHECK([ovs-appctl ofproto/trace br0 > > 'in_port=ovs-p0,dl_src=e4:11:22:33:44:55,dl_dst=e4:11:22:33:44:66,dl_type=0x0800,nw_src=10.1.1.1,nw_dst=10.1.1.12'], > > [0], [stdout]) > > +AT_CHECK([tail -3 stdout], [0], [dnl > > +Datapath actions: > > emit_sample(group=10,cookie=5555555566666666),userspace(pid=4294967295,flow_sample(probability=65535,collector_set_id=1,obs_domain_id=1431655765,obs_point_id=1717986918,output_port=4294967295)),trunc(100),3 > > +This flow is handled by the userspace slow path because it: > > + - Uses action(s) not supported by datapath. > > +]) > > + > > +OVS_DAEMONIZE([ovstest test-psample > psample_slow.out], > > [psample_slow.pid]) > > + > > +NS_CHECK_EXEC([at_ns0], [ping -q -c 1 10.1.1.2 | FORMAT_PING], [0], [dnl > > +1 packets transmitted, 1 received, 0% packet loss, time 0ms > > +]) > > + > > +AT_CHECK([grep -E 'SAMPLE1' psample_slow.out >/dev/null]) > > +AT_CHECK([grep -E 'SAMPLE2' psample_slow.out >/dev/null]) > > + > > +AT_CHECK([ovs-ofctl dump-ipfix-flow br0 | sed 's/tx pkts=[[0-9]]*/tx > > pkts=24/' | sed 's/tx errs=[[0-9]]*/tx errs=0/'], [0], [dnl > > +NXST_IPFIX_FLOW reply (xid=0x2): 2 ids > > + id 1: flows=2, current flows=0, sampled pkts=2, ipv4 ok=2, ipv6 ok=0, > > tx pkts=24 > > + pkts errs=0, ipv4 errs=0, ipv6 errs=0, tx errs=0 > > + id 2: flows=2, current flows=0, sampled pkts=2, ipv4 ok=2, ipv6 ok=0, > > tx pkts=24 > > + pkts errs=0, ipv4 errs=0, ipv6 errs=0, tx errs=0 > > +]) > > + > > +AT_CHECK([ovs-appctl lsample/show br0], [0], [dnl > > +local sample statistics for bridge "br0": > > +- Collector Set ID: 1 > > + Local Sample Group: 10 > > + Total number of bytes: 196 > > + Total number of packets: 2 > > + > > +- Collector Set ID: 2 > > + Local Sample Group: 12 > > + Total number of bytes: 196 > > + Total number of packets: 2 > > +]) > > + > > +OVS_TRAFFIC_VSWITCHD_STOP(["/sending to collector failed/d"]) > > +AT_CLEANUP > > -- > > 2.45.1 > _______________________________________________ dev mailing list d...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-dev
