Pico W http server slow response

[gwynge]

I'm using a Pico W as a monitor with an http server running via asyncio.start_server.

Every so often http request goes from taking perhaps 200mS to taking 15+ seconds. It does eventually respond. I'm struggling to reproduce when connected to Thonny console which is a bit annoying.

Anybody seen this sort of behaviour and have any wisdom?

[n1kdo]

do you ever call gc.collect() in your program to force garbage collection to occur -- when you want it to, rather than when the system decides to do it?

[gwynge]

I looked at memory usage, it fills up and collects and couldn't make anything happen in that respect so suspect it's not related (to this particular behaviour anyway)

[fivdi]

Maybe the power-saving mode on the wireless chip which is activated by default is causing the server to be less responsive. For further information see section "3.6.3. Power-saving mode" of Connecting to the Internet with the raspberry Pi Pico W.

[gwynge]

I'm sure I'd read the documentation before but not seen any mention of power save mode, Wonder if I found an old document? It also defines the response code of 2 when connecting which I couldn't find a definition of anywhere else!
I shall try disabling powersave and see if that works. I wondered if there was such a feature. Although I note that "idle" is not defined. In other words how long before it goes into power save mode?

[n1kdo]

I can serve static html pages from Pico-W in < .002 sec., so something is not right.

Here is the no-powersave code that I use to start the network.

        print('Connecting to WLAN...')
        wlan = network.WLAN(network.STA_IF)
        wlan.config(pm=0xa11140)  # disable power save, this is a server.
        # wlan.config(hostname='not supported on pico-w')
        wlan.active(True)
        wlan.connect(ssid, secret)

[gwynge]

Thanks, I shall try this.

[DavesCodeMusings]

I'm assuming you're coding your own http server here and not using a library. If so, use netstat on the client side to check for tcp connections in a TIME_WAIT or FIN_WAIT state. I had a similar problem where requests were not ended properly and the http server was running out of connections. That was what caused the odd, intermittent hanging for me. I think the default number of connections with start_server is 5, so if you see 5 or more waiting in the netstat output, you know what the problem is.

Here's a link to the details of my original problem and its solution. #9695