Description
After completing a series of tests it's found out that PFC storm detection and restoration time violate configured values.
Describe the results you received:
25 iterations were performed for each pair of PFCWD storm detection/restoration time.
Below you can see the average value of PFC storm detection/restoration time.
The average values exceed configured ones:
(table 1)
Results3.xlsx
Describe the results you expected:
PFCWD reaction time should be within ( or so ) configured values.
Steps to reproduce the issue:
- Configure qos on DUT
config qos reload
- Configure priority 2 as a lossless one.
- Assign IP-addresses to DUT and the host.
- Configure PFCWD on DUT:
pfcwd counter_poll enable
pfcwd interval 100
pfcwd start ports all -a drop detection-time 200 -r 200
- Sync time between DUT and the host.
- Write in syslog of both DUT and host an iteration start time:
logger -p user.error -t PFCTEST "Iteration # $i started at date +"%T.%N" "
- Inject from the host PFC storm:
quanta=65535
prio=2
amount of PFC frames=999999
- Write in syslog of both DUT and host an iteration finish time:
logger -p user.error -t PFCTEST "Iteration # $i finished at date +"%T.%N"
- Check in DUT's syslog a time of detection of PFC storm (created at step#7) :
Jan 2 14:11:17.291499 sonic NOTICE swss#orchagent: :- startWdActionOnQueue: Receive notification, storm
- Check in DUT's syslog a time of restoration PFC storm:
Jan 2 14:11:22.663897 sonic NOTICE swss#orchagent: :- startWdActionOnQueue: Receive notification, restore
- Repeat steps 6-10 25 times.
- Using data from steps 6, 8-10 calculate PFC storm detection/restoration times.
- Calculate average detection/restoration time for performed 25 iterations. See table 1.
- Increase PFCWD detection/restoration time to 400 msec. on DUT:
pfcwd start ports all -a drop detection-time 400 -r 400
- Repeat steps 5-13
- Increase PFCWD detection/restoration time to 600 msec. on DUT:
pfcwd start ports all -a drop detection-time 600 -r 600
- Repeat steps 5-13
Output of show version:
SONiC Software Version: SONiC.HEAD.129-0c9040de
Distribution: Debian 9.11
Kernel: 4.9.0-9-2-amd64
Build commit: 0c9040de
Build date: Thu Nov 21 12:50:41 UTC 2019
Built by: johnar@jenkins-worker-4
Platform: x86_64-mlnx_msn2100-r0
HwSKU: ACS-MSN2100
ASIC: mellanox
Serial Number: MT1646X05314
Uptime: 14:44:36 up 5:05, 2 users, load average: 3.78, 3.93, 3.97
Docker images:
REPOSITORY TAG IMAGE ID SIZE
docker-syncd-mlnx HEAD.129-0c9040de 4f3b126274c2 373MB
docker-syncd-mlnx latest 4f3b126274c2 373MB
docker-fpm-frr HEAD.129-0c9040de 7345e105bf26 321MB
docker-fpm-frr latest 7345e105bf26 321MB
docker-sflow HEAD.129-0c9040de f240b81ea842 305MB
docker-sflow latest f240b81ea842 305MB
docker-lldp-sv2 HEAD.129-0c9040de f1b94915bf13 299MB
docker-lldp-sv2 latest f1b94915bf13 299MB
docker-dhcp-relay HEAD.129-0c9040de 7ba3cee353b6 289MB
docker-dhcp-relay latest 7ba3cee353b6 289MB
docker-database HEAD.129-0c9040de c5c92210277c 281MB
docker-database latest c5c92210277c 281MB
docker-snmp-sv2 HEAD.129-0c9040de ad01f8a547b1 335MB
docker-snmp-sv2 latest ad01f8a547b1 335MB
docker-orchagent HEAD.129-0c9040de 7fc54899dc92 322MB
docker-orchagent latest 7fc54899dc92 322MB
docker-teamd HEAD.129-0c9040de c0316e26bbbd 304MB
docker-teamd latest c0316e26bbbd 304MB
docker-sonic-telemetry HEAD.129-0c9040de 43df42f49e91 304MB
docker-sonic-telemetry latest 43df42f49e91 304MB
docker-router-advertiser HEAD.129-0c9040de b0aff9280842 281MB
docker-router-advertiser latest b0aff9280842 281MB
docker-platform-monitor HEAD.129-0c9040de 10013f82e4d7 565MB
docker-platform-monitor latest 10013f82e4d7 565MB
Attach debug file sudo generate_dump:
sonic_dump_r-ethernet-sw230_20200102_143347.tar.gz
Description
After completing a series of tests it's found out that PFC storm detection and restoration time violate configured values.
Describe the results you received:
25 iterations were performed for each pair of PFCWD storm detection/restoration time.
Below you can see the average value of PFC storm detection/restoration time.
The average values exceed configured ones:
(table 1)
Results3.xlsx
Describe the results you expected:
PFCWD reaction time should be within ( or so ) configured values.
Steps to reproduce the issue:
config qos reloadpfcwd counter_poll enablepfcwd interval 100pfcwd start ports all -a drop detection-time 200 -r 200logger -p user.error -t PFCTEST "Iteration # $i started atdate +"%T.%N""quanta=65535prio=2amount of PFC frames=999999logger -p user.error -t PFCTEST "Iteration # $i finished atdate +"%T.%N"Jan 2 14:11:17.291499 sonic NOTICE swss#orchagent: :- startWdActionOnQueue: Receive notification, stormJan 2 14:11:22.663897 sonic NOTICE swss#orchagent: :- startWdActionOnQueue: Receive notification, restorepfcwd start ports all -a drop detection-time 400 -r 400pfcwd start ports all -a drop detection-time 600 -r 600Output of
show version:Attach debug file
sudo generate_dump:sonic_dump_r-ethernet-sw230_20200102_143347.tar.gz