What sort of IO Pode is doing in the background

[nightroman]

I have a basic Pode server, nothing advanced (tried the bare minimum server, too). I've noticed IO (and slight CPU) spikes approximately every 10 seconds. Out of curiosity, for better understanding, what is Pode doing in the background?

[mdaneri]

For IO do you mean disk or network ?

[nightroman]

It's the Process Explorer IO graph. Per this answer, it's

I/O Read Bytes - The number of bytes read in input/output operations generated by a process, including file, network, and device I/Os. I/O Read Bytes directed to CONSOLE (console input object) handles are not counted.

[nightroman]

Other observations from Process Explorer. CPU consumption is constantly non-zero. Threads are constantly working, e.g. see highlighted below with non-zero CPU, their number change frequently (~1 sec) with 3 to 7 threads doing something any time.

[mdaneri]

There are multiple runspaces running. Why are you surprised?
You can see them using get-runspace inside a debugging session

[nightroman]

I am curious about what these workspaces are doing, for knowledge, understanding. Runspaces are not "running" just because they are created.

I am not sending any requests to Pode, it is effectively idle. But the amount of observed activity in the idle mode might be slightly surprising.

[nightroman]

@mdaneri I understand that some activity is probably needed by design. But can its intensity be somehow configured? In some future at least. For example as a configurable factor applied to all used Start-Sleep intervals used internally (assuming something is running in loops with sleeps).

My use case: the server might be needed at any moment, so it runs permanently (as a windows service).
But the reality is that moments of using it are rare, it might be used for just a dozen of requests a day.
Thus, I would prefer to configure the server to be "more idle", more sleepy, than it is now.

[nightroman]

Some quick digging shows that (in my case at least) logging and cache housekeeping are constantly doing some work even if the server is not actually used at all.

[nightroman]

Interestingly, the cache housekeeping is internally added as a timer, with its interval 10 sec., and it is the only timer I have (added by the system). But the timer checking loop is running with Start-Sleep 1, i.e. in my case it is probably waking up "too frequently".

[Badgerati]

Outside of the threading for the socket listener, Pode does create various Timers for housekeeping functionality. Some, as you've seen like the Logging/Caching, are created by Pode all the time; while others, like Sessions and Tasks, are only spawned if you actually configure them.

We could add in configuration to allow for tweaking the internal Timer intervals, but it would have to come with a big WARNING label - as making them too short/long could come with quite adverse affects.

Another approach, for the Timers and for the Timer Checker, is to just make them more "intelligent" when there's less to do:

• For the Timers we could use some form of capped exponential/linear back-off, and each time the logic fully triggers the back-off is reset - for logging maybe something like 100ms > 500ms > 5s > 20s > 1m, and if it has to log something we start at 100ms again. This would make them more "idle", and we have better control of sensible idleness for certain timers.
• For the Timer Check, we could probably make it sleep for a delta time rather than always 1s. So figure out when the next Timer is, and sleep for that long 🤔 this might will be harder to implement, but should be possible.

[nightroman]

I think the approach with Start-Sleep is too simple (and not effective) for such a complex engine as Pode.
I would explore some different and several alternative. E.g. something like quick and dirty POC below.

$wait = [System.Threading.AutoResetEvent]::new($false)

# These two parallel loops fake some events.
1..2 | %{
	$ps = [powershell]::Create()
	$null = $ps.AddScript({
		param($wait)
		while($true) {
			Start-Sleep (Get-Random 10)
			$wait.Set()
		}
	})
	$null = $ps.AddArgument($wait).InvokeAsync()
}

# This loop does not know for how long to sleep.
# It wakes up as soon as relevant events happen.
# (and relevant events call handle.Set())
while($true) {
	Write-Host "Sleeping..."
	$sw = [System.Diagnostics.Stopwatch]::StartNew()

	# sleep as long as needed
	$null = $wait.WaitOne()
	Write-Host "Resuming after $($sw.Elapsed)"
}

[nightroman]

Some thoughts about default cache housekeeping. The hardcoded 10 seconds timer interval seems to be too aggressive, given

The default TTL for cached items is 3,600 seconds (1 hour), and this value can be customised either globally or per item.
There is also a $cache: scoped variable available for use.

Thus, for 1 created cache item with default expiration, the housekeeping wakes up and performs its job (lock, clone keys, check keys) 360 times before actually collecting the expired item.

Besides, unless I miss something, the housekeeping does not have to remove expired items "ASAP". There is nothing wrong with some expired items staying in memory reasonably longer.

This may be easily addressed if the housekeeping interval is customizable (together with the default TTL). Those who care about this sort of issues, may set it to the appropriate value.

[nightroman]

Is there a straightforward way to opt-out of default caching? I would like to avoid its activity because caching is not used.

I have found one peculiar way to work around the issue. Adding a dummy cache storage does the trick:

Add-PodeCacheStorage -Name None -Get {} -Set {} -Remove {} -Test {} -Clear {} -Default

Ideally, instead of empty script blocks some { throw "Cache not supported" } should be used.

The main issue of unnecessary CPU activity and IO spikes is resolved. The logging loop is still doing something but it is less intrusive (and at least logging is used occasionally).