proxy and worker

Hi all,
I have an issue about proxy and worker.Here is my cluster configuration:

[manager]
type=manager
host=localhost

[proxy-1]
type=proxy
host=localhost

[worker-1]
type=worker
host=localhost
interface=ens512f1
lb_method=pf_ring
lb_procs=8
pin_cpus=0,1,2,3,4,5,6,7

[worker-2]
type=worker
host=localhost
interface=ens512f1
lb_method=pf_ring
lb_procs=8
pin_cpus=8,9,10,11,12,13,14,15

[worker-3]
type=worker
host=localhost
interface=ens512f1
lb_method=pf_ring
lb_procs=8
pin_cpus=16,17,18,19,20,21,22,23

[worker-4]
type=worker
host=localhost
interface=ens512f1
lb_method=pf_ring
lb_procs=4
pin_cpus=24,25,26,27,28,29,30,31

If I run all threads on single worker like this:

[worker]
type=worker
host=localhost
interface=ens512f1
lb_method=pf_ring
lb_procs=32
pin_cpus=0,1,2,3,4,5,6,…31

What’s the difference between all threads on a single worker vs threads on multiple workers?

When I use threads on multiple workers on a single server, is there any difference one and multiple proxy?

Any insight would be helpful.

Xu Guo

Hi all,
    I have an issue about proxy and worker.Here is my cluster configuration:

[manager]
type=manager
host=localhost

[proxy-1]
type=proxy
host=localhost

[worker-1]
type=worker
host=localhost
interface=ens512f1
lb_method=pf_ring
lb_procs=8
pin_cpus=0,1,2,3,4,5,6,7

[worker-2]
type=worker
host=localhost
interface=ens512f1
lb_method=pf_ring
lb_procs=8
pin_cpus=8,9,10,11,12,13,14,15

[worker-3]
type=worker
host=localhost
interface=ens512f1
lb_method=pf_ring
lb_procs=8
pin_cpus=16,17,18,19,20,21,22,23

[worker-4]
type=worker
host=localhost
interface=ens512f1
lb_method=pf_ring
lb_procs=4
pin_cpus=24,25,26,27,28,29,30,31

The above configuration is not valid.

If I run all threads on single worker like this:

[worker]
type=worker
host=localhost
interface=ens512f1
lb_method=pf_ring
lb_procs=32
pin_cpus=0,1,2,3,4,5,6,....31

What's the difference between all threads on a single worker vs threads on multiple workers?

As the "procs" in the option indicates, lb_procs creates multiple processes, not threads.

The first configuration creates 4 groups of processes that will all see duplicates of the same traffic.

You want the 2nd configuration.

You also likely do not have a real 32 core system. If that is 32 cores with hyper threading or something, you only want to run about 15 workers.