Netgate 8200 what tests to do with iPerf3 -> problems

SwissSteph · Jun 10, 2024, 7:52 AM

I've had a Netgate 8200 for about a year and on my desktop PC I've switched from an "Asus XG-C100C" card (10 Gb) to an "Intel x550-T2" (10 Gb).

While with the Asus I was able to run iPerfs tets without any weird problems, the same tests with the Intel card blocked my entire LAN (and therefore all the devices connected to it, including those in VLAN).

Only my connection, with another computer (laptop), on another port of the 8200 (which I named "Emergency" - igc4) could connect to the WEB, but logically not at all to my LAN or my other VLANs.

The only solution I found (after a long moment of panic) was to simply restart my pfsense to get back to a normal situation where everything was working (phew).

I can reproduce the blockage without any problem, I run different iPerf3 commands from my desktop PC and pfsense is completely blocked.

Is it a normal situation to be able to so easily overload an 8200 to such a blocking point?

Thank you for your advice and analysis

stephenw10 · Jun 10, 2024, 12:50 PM

Do you mean while the test is running? Or that it still fails after the test has finished?

Do you see anything logged? Anything on the Interfaces > Status page? Is the NIC still showing as linked?

Steve

SwissSteph · Jun 10, 2024, 1:51 PM

Hello Steve,

I'll put the parameters I used to run my tests here later.

In fact, I've “noticed” one thing... if I run (in a CMD” command) every test BUT waiting for pfsense's CPU and memory usage to come down, I have a harder time blocking everything.

Typically, I run the “iperf3” command and in the best case I get 2 or 3 lines displayed, then nothing in the CMD window, iperf seems frozen (I have to press ctrl-C to stop the process).

Then my PC is still connected (GUI) to pfsense and fortunately I can still access the “reboot” command... in the meantime, I have no wifi, no access to my switches and others (which are on my LAN like my VLANs).

I'm sorry my explanations aren't always very clear... my English is just Google translated English.

I'm currently reading the https://forum.netgate.com/topic/182534/just-purchased-a-netgate-8200-having-a-few-issues/12 topic and I've just put in the proposed modifications. I'll wait until I get home to run a new test with these values and see if I freeze pfsense again.

EDIT
With the modifications made (thanks) on the link above, here are the results AND, above all, I was no longer able to block pfsense (just slow it down a little).

iperf3.exe -c 192.168.1.1 -P 100

[SUM] 0.00-10.00 sec 8.30 GBytes 7.13 Gbits/sec sender
[SUM] 0.00-10.00 sec 8.27 GBytes 7.11 Gbits/sec receiver

iperf3.exe -c 192.168.1.1 -P 10

[SUM] 0.00-10.00 sec 5.69 GBytes 4.89 Gbits/sec sender
[SUM] 0.00-10.00 sec 5.69 GBytes 4.88 Gbits/sec receiver

iperf3.exe -c 192.168.1.1 -P 50

[SUM] 0.00-10.00 sec 8.14 GBytes 6.99 Gbits/sec sender
[SUM] 0.00-10.00 sec 8.13 GBytes 6.99 Gbits/sec receiver

iperf3.exe -c 192.168.1.1 -P 8

[SUM] 0.00-10.00 sec 5.47 GBytes 4.70 Gbits/sec sender
[SUM] 0.00-10.00 sec 5.47 GBytes 4.70 Gbits/sec receiver

vmstat -i
cpu0:timer 438153 208
cpu1:timer 851223 403
cpu2:timer 442786 210
cpu3:timer 551482 261
cpu4:timer 331664 157
cpu5:timer 434263 206
cpu6:timer 517970 245
cpu7:timer 305809 145

stephenw10 · Jun 10, 2024, 2:19 PM

There is little point using more than 8 parallel streams because the ix NICs have 8 queues.

How exactly are you testing? From a Windows client to pfSense directly as the server?

SwissSteph · Jun 10, 2024, 4:21 PM

Thanks for clarifying the “8” parameter

Yes, Pfsense iPerf server and on my Windows 11 I run iPerf

I currently have these settings:

net.isr.maxthreads="-1"
net.isr.bindthreads="1"
machdep.hyperthreading_intr_allowed="1"

legal.intel_iwi.license_ack="1"
kern.ipc.nmbclusters="1000000"
kern.ipc.nmbjumbop="524288"
dev.ix.0.iflib.override_nrxqs="8"
dev.ix.0.iflib.override_ntxqs="8"
dev.ix.1.iflib.override_nrxqs="8"
dev.ix.1.iflib.override_ntxqs="8"
dev.ix.0.iflib.separate_txrx="1"
dev.ix.1.iflib.separate_txrx="1"
dev.ix.0.iflib.rx_budget="65535"
dev.ix.1.iflib.rx_budget="65535"
dev.ix.0.iflib.override_nrxds="4096"
dev.ix.1.iflib.override_nrxds="4096"
dev.ix.0.iflib.override_ntxds="4096"
dev.ix.1.iflib.override_ntxds="4096"

hw.intr_storm_threshold="10000"

What do you think of these “numbers”?
I read somewhere that you should double the value set for “Receive buffers” (here 4096) compared with the value for “Transmit buffers” (here 4096).

Should I set “8192” for “ntxds”?

My speed directly from my FAI box

My speed from my PC (Windows 11) through my 8200

stephenw10 · Jun 10, 2024, 6:52 PM

What does the CPU usage across the cores look like whilst that test is running?

SwissSteph · Jun 10, 2024, 7:09 PM

@stephenw10
Which command should I run?

top -aSH

stephenw10 · Jun 10, 2024, 8:13 PM

The full output of top -HaSP will show the per core loading like:

last pid: 13979;  load averages:    0.15,    0.22,    0.21               up 0+08:08:55  21:13:08
307 threads:   5 running, 285 sleeping, 17 waiting
CPU 0:  0.0% user,  0.0% nice,  0.8% system,  0.0% interrupt, 99.2% idle
CPU 1:  0.0% user,  0.4% nice,  3.2% system,  0.0% interrupt, 96.4% idle
CPU 2:  0.0% user,  0.8% nice,  1.2% system,  0.0% interrupt, 98.0% idle
CPU 3:  0.0% user,  1.2% nice,  2.4% system,  0.0% interrupt, 96.5% idle
Mem: 44M Active, 226M Inact, 378M Wired, 56K Buf, 3067M Free
ARC: 152M Total, 36M MFU, 112M MRU, 128K Anon, 960K Header, 4201K Other
     119M Compressed, 276M Uncompressed, 2.32:1 Ratio
Swap: 1024M Total, 1024M Free

  PID USERNAME    PRI NICE   SIZE    RES STATE    C   TIME    WCPU COMMAND
   11 root        187 ki31     0B    64K RUN      0 474:39  97.43% [idle{idle: cpu0}]
   11 root        187 ki31     0B    64K CPU3     3 475:37  96.43% [idle{idle: cpu3}]
   11 root        187 ki31     0B    64K CPU1     1 474:52  95.39% [idle{idle: cpu1}]
   11 root        187 ki31     0B    64K CPU2     2 474:54  95.19% [idle{idle: cpu2}]
 6150 root         68   20    13M  3076K piperd   2   0:04   0.24% /bin/sh /var/db/rrd/updaterrd
    0 root        -12    -     0B  1664K -        0   0:10   0.20% [kernel{z_wr_iss_0}]

SwissSteph · Jun 11, 2024, 1:38 PM

@stephenw10

iperf3.exe -P 8 -c 192.168.1.1

two tests just in

stephenw10 · Jun 11, 2024, 1:42 PM

Ok so some CPU cores there are running at 100%. I would try testing with ntopng disabled since that's using 77% of one core. But it's unlikely you will get much faster than that.

SwissSteph · Jun 11, 2024, 1:51 PM

@stephenw10

Thank you very much for your analysis and advice!

With the configuration changes mentioned above, I no longer have pfsense blocking, it's a bit of a shame that some settings aren't more “original” configured for a (in my case) 8200.

I'm glad to have found the https://forum.netgate.com/topic/182534/just-purchased-a-netgate-8200-having-a-few-issues/13topic which helped me enormously to find a solution to my problem.

EDIT

Last test