[Win10 2020.22080] WebClientDAT freezes Process

[alphamoonbase]

Im working alot with the WebClientDAT and have come to a point where it is unusable in this state, sadly.
At first I thought it had to do with all my scripting I do on receive, but this was a dead end, so I tested a vanilla setup and the freeze continued.
I talked to markus and he suggested that disabling the textoutput might be a solution and this, at least, drasticly reduced the cooktime of the WebClientDAT, but the Freeze continued (even with low cook times of the webClientDAT.

So I made a testsetup and realized that the freeze is indipendet of the cooking of the DAT but appears somewhere in between and not always at the same place.
Cook_time and cooked_this_frame come from the WebClientDAT, FPS and msec are from the overall project performance.
received_response and connect are triggered on the callbacks.

webclientTestbench.toe (4.1 KB)

Edit: Also tested on another mashine running 22080 and have the same problem, but less harsh, only double the time:

[MXZEHN]

Just run your testbench.toe and found something rather interresting. The first time after sending a request I experience the same issue of the high frame time

But after that the hiogh frametimes do no longer happen.

Also using 22080 here

[eric.b]

I tried your toe file but no luck reproducing. Are you just sending 1 request at a time to derivative.ca via the Request pulse parameter?

[alphamoonbase]

Hoi Eric. Thanks for trying.
Im using just the standart configuration except killing the textOutput to get a clearer reading and sending only one request at a time.

Trying it out right now again and it seems that im getting actually two spikes now:

Something interresting might also be that disconnecting the connection via the stop parameter during the receiving results in a similiar spike.

Also something of interrest might be that the spike is not always directly after the request but sometimes inbetween:

Sometimes, im not 100% Sure if this is actually the case, it seems that the spike does not happen but its not reproducable.

Right now I testes 3 Setups:

• My Dev PC with Win10 Pro with all Updates an an old i5 6600 with lots of other Software running
• A Laptop Win10 Pro with all Updates with lots of stuff running where the spike only occures the first time
• A Production mashine Running LTSC completly virgin also showing the same spikes, but to lesser extend.
• Another Laptop same result

Edit:
Now the spike vanished. I only paused the timeline and wrote the part above. Now the spike is gone.

After a restart the spike is back.

[alphamoonbase]

Just a quick update.
Im able to reproduce a way to get rid of the spike.
When I switch to another Program (in this case my browser) for >30s the spike vanishes.
So weird, gonna do some more testing to give you a clearer image!

[eric.b]

I tried on another computer and I’m able to reproduce the issue, not quite as drastic it seems though – I get spikes up to about 35ms. It’s not immediately clear where that spike is coming from so I’ll take a closer look.