Hello Everyone,
Testing zeek 4.0.2 code on centos 7. Workers on 4.0.2 code are running with really high CPU compared to 3.0.13 code even when there is no traffic on the myricom interface.
Compiled zeek 4.0.2 with following options:
./configure --spooldir=/var/spool/zeek --logdir=/var/log/zeek --with-pcap=/opt/snf/ --enable-perftools --disable-zkg --disable-broker-tests --disable-btest --disable-archiver
Version of myricom driver we are running:
Product Code | Version |
- | - |
10G-PCIE2-8C2-2S | 3.0.25.50927
|
CPU utilization jumps very very high when we load balance worker processes compared to a single process on 4.0.2 code.
node.cfg with couples of processes and top output:
[logger]
type=logger
host=localhost
pin_cpus=16
[manager]
type=manager
host=localhost
pin_cpus=14
[proxy-1]
type=proxy
host=localhost
pin_cpus=12
[worker-1]
type=worker
host=localhost
Env_Vars = SNF_APP_ID=3, SNF_NUM_RINGS=2, SNF_RSS_FLAGS=0x31, SNF_DATARING_SIZE=8192MB, SNF_DESCRING_SIZE=2048MB, SNF_FLAGS=0x1, SNF_DEBUG_MASK=3
lb_method=myricom
lb_procs=2
pin_cpus=8,10
interface=snf0
Top output:
From 4.0.2 code:
sudo -u zeek /usr/local/zeek/bin/zeekctl top
Name Type Host Pid VSize Rss Cpu Cmd
logger logger localhost 20805 255M 79M 0% zeek
manager manager localhost 20856 216M 77M 0% zeek
proxy-1 proxy localhost 20906 215M 77M 0% zeek
worker-1-1 worker localhost 20966 20G 20G 70% zeek
worker-1-2 worker localhost 20971 20G 20G 70% zeek
From 3.0.13 code:
Name Type Host Pid VSize Rss Cpu Cmd
logger logger localhost 11179 254M 85M 0% zeek
manager manager localhost 11227 207M 86M 12% zeek
proxy-1 proxy localhost 11278 206M 87M 6% zeek
worker-1-1 worker localhost 11339 20G 20G 6% zeek
worker-1-2 worker localhost 11341 20G 20G 12% zeek
node.cfg with single process and top output:
[logger]
type=logger
host=localhost
pin_cpus=16
[manager]
type=manager
host=localhost
pin_cpus=14
[proxy-1]
type=proxy
host=localhost
pin_cpus=12
[worker-1]
type=worker
host=localhost
Env_Vars = SNF_APP_ID=3, SNF_NUM_RINGS=1, SNF_RSS_FLAGS=0x31, SNF_DATARING_SIZE=8192MB, SNF_DESCRING_SIZE=2048MB, SNF_FLAGS=0x1, SNF_DEBUG_MASK=3
pin_cpus=8
interface=snf0
Top output:
From 4.0.2 code:
Name Type Host Pid VSize Rss Cpu Cmd
logger logger localhost 24484 255M 76M 0% zeek
manager manager localhost 24535 216M 80M 0% zeek
proxy-1 proxy localhost 24586 215M 76M 0% zeek
worker-1 worker localhost 24635 20G 20G 18% zeek
From 3.0.13 code:
Name Type Host Pid VSize Rss Cpu Cmd
logger logger localhost 39901 245M 83M 5% zeek
manager manager localhost 39949 206M 86M 0% zeek
proxy-1 proxy localhost 39996 205M 83M 5% zeek
worker-1 worker localhost 40043 20G 20G 11% zeek
Has anyone run into this issue? Am I misconfiguring something on zeek 4.0.2?