The deployment is a follows:
- Cluster has with two nodes, each with 2 workers and the workers are pinned to specific cpu cores.
- dell r620 and centos7 with PF_RING
[root@centos soft]# uname -a
Linux centos 3.10.0-327.13.1.el7.x86_64 #1 SMP Thu Mar 31 16:04:38 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux
[root@centos~]# cat /proc/net/pf_ring/info
PF_RING Version : 6.3.0 (unknown)
Total rings : 4
Standard (non DNA/ZC) Options
Ring slots : 32768
Slot version : 16
Capture TX : Yes [RX+TX]
IP Defragment : No
Socket Mode : Standard
Total plugins : 0
Cluster Fragment Queue : 0
Cluster Fragment Discard : 0
broctl capture
[BroControl] > netstats
worker-1-1: 1460663117.341926 recvd=5317 dropped=0 link=5317
worker-1-2: 1460663117.542621 recvd=524 dropped=0 link=524
worker-2-1: 1460663117.751165 recvd=17 dropped=0 link=17
worker-2-2: 1460663117.945068 recvd=14417 dropped=0 link=14417
You can see almost all of the traffic capture by the woker-2-2.
And I downloaded 100 files, and only 25 files that could be extracted.
I found an very interesting phenomenon。
If I start 1 workers, all of the files can be extracted.
If I start 2 workers, 50% of the files can be extracted.
If I start 4 workers, 25% of the files can be extracted.
…
I don’t know why .
Could any one give me some advice.Thanks very much.