PS6 Windows memory leak crash

bboyes · Post by **bboyes** » Thu Apr 25, 2019 10:43 pm

Yesterday PS6 on Win10 Pro crashed after not even two days. Here are the key lines from trace.xml
22/04/19 07:48:56.523" category="Interaction" level="Info">Run.Pressed=True
24/04/19 01:17:37.487" category="Exception" level="Error">OutOfMemoryException handled in DRQ:System.OutOfMemoryException: Exception of type 'System.OutOfMemoryException' was thrown.
at G..ctor(w deviceModel, H type, UInt32 size, d indices)
at fl.s.a(w , Int32 , d )
at fl.DirectCollectProgressiveData(w deviceModel, Int32 numberOfSamples, d indeces)
at ev.Pico.Common.Dal.Device.IDataRequestQueueClient.CollectData(G& data, w deviceModel, UInt16 segmentIndex, C cancel, c dataRequest)
at eK.a.a()

I submitted a topic here after the crash on 22 April but it has not been "approved" or listed here so I have no way to know if anyone received it. So I am trying again. We really would like to have this fixed. I also sent an email with the whole log attached, to support@, more details in that.
Sincerely...

Post by **Martyn** » Fri Apr 26, 2019 5:41 am

We received the email, and have been trying to replicate the issue, but without success. We will follow up through the ticketing system.

bboyes · Post by **bboyes** » Fri Apr 26, 2019 11:06 am

Thank you, I will run the interim release today. Installing it now. If I run task manager with the current stable version, I can actually see memory going away - after less than 24 hours, memory use has increased from 34% to 64%. I assume you have tested with some memory use monitor running. It seems to lose a bit every time it redraws the screen, so if that is only every time we have a capture, that would be consistent with what we see in the field vs in the office with a much faster rate of triggered captures. I will keep you posted.

terrypin · Post by **terrypin** » Sun Apr 28, 2019 3:23 pm

As I’ve had many recent crashes of Picoscope 6.11.10 on my XP Pro PC, this thread prompted me to look at its Mem Usage in TM. Overnight in Stopped mode that increased from 124 MB TO 235 MB. Is that unexpectedly high? Why should it increase significantly at all if not collecting samples? Perhaps this is not only a Win 10 problem? Although I’d have expected to have seen more posts about it....

Terry, East Grinstead, UK

bboyes · Post by **bboyes** » Mon Apr 29, 2019 5:36 am

Memory leaks are typically in the program, not the OS. The application memory usage should not grow over time if its settings are not changed.

bboyes · Post by **bboyes** » Mon Apr 29, 2019 5:43 am

OK we now have two unstable PS6 systems. I had to visit the remote site on Saturday as both systems appeared to be no longer connected to the cloud. Since I had only installed the new PS4444 in place Friday noon-ish I assumed the issue was the cloud connection.
That assumption was incorrect. Here is what I found:
PS4444 with interim PS6 release had what looked like a functioning program display but nothing actually worked. It was locked up. I had to terminate it with Task Manager, zip the logs, and restart it. Logs attached.So this one failed after not even running 12 hours.
PS3425 with the current public PS6 release had terminated. I zipped its logs and restarted it. This one had run for several days prior to failing. Logs attached.
I have not had the time to look at these log files: maybe tomorrow. I have my hands full with other project issues.
Sincerely
Bruce Boyes

bboyes · Post by **bboyes** » Mon Apr 29, 2019 3:27 pm

Looking at trace.xml from the PS4444, I find this odd bit:
BackgroundHttpClientRequester failing with The underlying connection was closed: The connection was closed unexpectedly.

What is PS6 trying to do with http?

I believe this scope made a couple of captures after these http failures; the last one was timestamped 14:53:55.

According to trace.xml the application was at least partly running when I arrived Sat evening. But it was unresponsive: could not be stopped, closed, zoomed or anything: GUI frozen.

bboyes · Post by **bboyes** » Mon Apr 29, 2019 3:48 pm

Looking at the trace.xml file from the PS3425 running the

Power change broadcast: Mains
BackgroundHttpClientRequester failing with Authentication failed because the remote party has closed the transport stream.
BackgroundHttpClientRequester failing with A task was canceled.
BackgroundHttpClientRequester failing with Authentication failed because the remote party has closed the transport stream.
BackgroundHttpClientRequester failing with Authentication failed because the remote party has closed the transport stream.
BackgroundHttpClientRequester failing with Authentication failed because the remote party has closed the transport stream.
The application was unexpectedly closed.

bboyes · Post by **bboyes** » Mon Apr 29, 2019 3:52 pm

Looking at the trace.xml file from the PS3425 running the public PS6 version, all seems to be well from Mar 19 until Apr 26 where failed http attempts from PS6 seem to trigger a crash:

Power change broadcast: Mains
BackgroundHttpClientRequester failing with Authentication failed because the remote party has closed the transport stream.
BackgroundHttpClientRequester failing with A task was canceled.
BackgroundHttpClientRequester failing with Authentication failed because the remote party has closed the transport stream.
BackgroundHttpClientRequester failing with Authentication failed because the remote party has closed the transport stream.
BackgroundHttpClientRequester failing with Authentication failed because the remote party has closed the transport stream.
The application was unexpectedly closed.

This system is on a metered connection via 4GLTE modem to get data to the cloud from a remote site adjacent railroad tracks and high voltage mains transmission lines.

Can anyone shed some light on what is going on here and how to prevent these crashes?

Post by **Martyn** » Tue Apr 30, 2019 6:07 am

The http messages are just warnings that come in soon after starting the software and cause no problem.

There are over 6 hours between the last http message and the "unexpectedly closed" message, which is actually written to the trace file by default during operation so that if there is a crash and the file write can't be completed you get that as the last line.

Trying to guess what all the messages mean is not straightforward, and without knowledge of the code can lead you down the wrong path.

I will discuss the issues with the development team.

Post by **Martyn** » Tue Apr 30, 2019 8:06 am

It would be worth checking in the Windows Event Viewer if there are any system actions occurring at around the same time as the timestamp of the last line in the trace.xml file

bboyes · Post by **bboyes** » Tue Apr 30, 2019 4:25 pm

Will check Windows event logs. Every trip to this site is now 2 hrs and it's a tedious process getting there and working in physically awkward circumstances. I have to go back today as apparently both systems have gone off line again. I was just there less than 12 hours ago... I have reverted the PS4444 system to 6.13.15. Yesterday AM both systems were offline so in addition to the reverting I set one trigger lower. That was a mistake. The rest of the day I blew through 12 GB of metered data -- but at least both were logging. I could not get back onsite until late at night and increase the trigger so I don't catch so much uninteresting data. This afternoon we are scheduled to have thunderstorms move in so want to have both systems online for that.

terrypin · Post by **terrypin** » Wed May 01, 2019 10:20 am

bboyes wrote: ↑
Mon Apr 29, 2019 5:36 am
Memory leaks are typically in the program, not the OS. The application memory usage should not grow over time if its settings are not changed.

Yes, I was (am) seeking a response from PicoTech on whether the PicoScope 6 problem was relevant to another OS than Win 10, such as my XP Pro (SP3).

Terry, East Grinstead, UK

bboyes · Post by **bboyes** » Wed May 01, 2019 12:38 pm

If I observe trace.xml remotely and don't see an unexpected termination entry does that mean the app should still be working? See my new post (still not approved for display) about the "zombie lockup" we are seeing.

Post by **Martyn** » Wed May 01, 2019 1:29 pm

Yes that is correct the application should still be running unless it has completely locked up and can't write to the trace file. The line is removed if the application is closed cleanly.

That line is in the trace file to report that the application has stopped unexpectedly, and not in a clean way, and is generally indicative of an issue between the application and the O/S

Test & Measurement Forum

PS6 Windows memory leak crash

PS6 Windows memory leak crash

Re: PS6 Windows memory leak crash

Re: PS6 Windows memory leak crash

Re: PS6 Windows memory leak crash

Re: PS6 Windows memory leak crash

Re: PS6 Windows memory leak crash

Re: PS6 Windows memory leak crash

Re: PS6 Windows memory leak crash

Re: PS6 Windows memory leak crash

Re: PS6 Windows memory leak crash

Re: PS6 Windows memory leak crash

Re: PS6 Windows memory leak crash

Re: PS6 Windows memory leak crash

Re: PS6 Windows memory leak crash

Re: PS6 Windows memory leak crash