K4ktus
Posts: 20
Joined: Sun Jun 30, 2019 5:22 pm

Raspberry Pi Diagnostics feedback

Thu Mar 05, 2020 3:13 pm

I wanted to share my experiences with recently released diagnostic tool. I'm running a Pi 3 B+ with current Raspbian install. First of all - when I first run the program and click "Run Tests", it does absolutely nothing, except saying "Not run" next to the SD card test option. Only when I click "Reset" and then "Run Tests" agin, it actually does the test. Secondly - My Pi is blazing fast, and yet it fails almost every test with terrible results, even as low as 0 (!). I checked dmesg for SD card errors and there were none. How should I interpret this? I tried --debug=io option suggested by the log, but after a while the test just aborted. Last but not least, the application freezes very often and ocasionally throws theme errors in the log.
I hope this helps developers improve the diagnostic tool, because I think it's an awesome concept, it just needs a bit of work. Of course I can provide my config or test log files, as well as conduct suggested tests on my own Pi.
Using Pi as a full-time desktop PC since January 2019

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 26408
Joined: Sat Jul 30, 2011 7:41 pm

Re: Raspberry Pi Diagnostics feedback

Thu Mar 05, 2020 4:43 pm

Pi can be fast, but SD card can still be slow, but it does sound you have other issues. I've run this today and it was fine, no locks, worked on first button press etc.

Is your OS installation all up to date?
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed.
I've been saying "Mucho" to my Spanish friend a lot more lately. It means a lot to him.

hippy
Posts: 7378
Joined: Fri Sep 09, 2011 10:34 pm
Location: UK

Re: Raspberry Pi Diagnostics feedback

Thu Mar 05, 2020 4:52 pm

K4ktus wrote:
Thu Mar 05, 2020 3:13 pm
when I first run the program and click "Run Tests", it does absolutely nothing, except saying "Not run" next to the SD card test option. Only when I click "Reset" and then "Run Tests" agin, it actually does the test.
Same for me, Pi 3B (non-plus) and other issues ... viewtopic.php?f=28&t=267094

User avatar
davidcoton
Posts: 4883
Joined: Mon Sep 01, 2014 2:37 pm
Location: Cambridge, UK
Contact: Website

Re: Raspberry Pi Diagnostics feedback

Thu Mar 05, 2020 4:54 pm

I can confirm the problem on first run, as described above (Pi4B 2GB, Sandisk Ultra Class 10 A1,fully updated Raspbian Buster).

Admittedly my card is not newly formatted, but on the second attempt (after reset) I got a confusing log (Pass 2 happened twice, four passes listed) with mixed results (each test passed at least once and failed at least once). Exited and re-run, one pass, everything passed.

So certainly there is a minor problem on first run, and results are inconsistent (which may be a real, justified observation, of course).

Can do more testing if it helps.
Signature retired

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 26408
Joined: Sat Jul 30, 2011 7:41 pm

Re: Raspberry Pi Diagnostics feedback

Thu Mar 05, 2020 5:07 pm

Have passed this thread on to the engineer involved. Not sure there is a github repo yet for official bug reports.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed.
I've been saying "Mucho" to my Spanish friend a lot more lately. It means a lot to him.

User avatar
TideMan
Posts: 259
Joined: Fri Jun 22, 2012 8:08 am
Location: ChCh, NZ

Re: Raspberry Pi Diagnostics feedback

Thu Mar 05, 2020 7:22 pm

Program works fine for me, but the results are a bit perplexing.
All the cards I've tested are PASS for reading, but FAIL for writing.
This includes an expensive MLC card.
Yet I have no perceptible problems writing to these cards.

So, what is the point of the tests?
If you wish in this world to advance, your merits you're bound to enhance.
You must stir it and stump it and blow your own trumpet, or trust me you haven't a chance.
Ruddigore, G&S

K4ktus
Posts: 20
Joined: Sun Jun 30, 2019 5:22 pm

Re: Raspberry Pi Diagnostics feedback

Thu Mar 05, 2020 8:38 pm

Is your OS installation all up to date?
Yes, I updated it right afrer installing the tool.
The SD card was not formated for quite a while, so doing that probably could improve the results. I'll try that and post an update here.
it does sound you have other issues
Could you explain a bit more? What kind of issues? How to deal with them?
Using Pi as a full-time desktop PC since January 2019

User avatar
Milliways
Posts: 534
Joined: Fri Apr 25, 2014 12:18 am
Location: Sydney, Australia

Re: Raspberry Pi Diagnostics feedback

Fri Mar 06, 2020 1:17 am

Tried to run on Raspbian Buster Lite

It started but exited with error

Code: Select all

fio: job startup hung? exiting.

MarkDH102
Posts: 402
Joined: Fri Feb 13, 2015 3:18 pm

Re: Raspberry Pi Diagnostics feedback

Fri Mar 06, 2020 6:57 am

I have tried this tool on 2 completely up to date Buster Pi's. One a Pi4 4GB and the other a Pi2B.
Both have Sandisk Ultra 16GB HC I class 10 cards. Both have the same base kernel on them.
The tool ran first time with no problems on both.
However the results are completely different. Governed no doubt by the underlying hardware and not the card.
Now, I've had the cards for a while and they've had good use but they were recently (this year) overwritten with the Buster Image using Win32diskimager.
OpenCVBusterPi4.png
OpenCVBusterPi4.png (122.1 KiB) Viewed 1550 times
OpenCVBusterPi2.png
OpenCVBusterPi2.png (130.31 KiB) Viewed 1550 times

hippy
Posts: 7378
Joined: Fri Sep 09, 2011 10:34 pm
Location: UK

Re: Raspberry Pi Diagnostics feedback

Fri Mar 06, 2020 12:48 pm

TideMan wrote:
Thu Mar 05, 2020 7:22 pm
So, what is the point of the tests?
To test SD Card reading and writing speeds to verify the card is appropriate for use with a Pi desktop system.

That the test and/or test harness is buggy, broken, or simply doesn't work for some, does not invalidate the point of it.

K4ktus
Posts: 20
Joined: Sun Jun 30, 2019 5:22 pm

Re: Raspberry Pi Diagnostics feedback

Fri Mar 06, 2020 9:00 pm

I formatted my SD card. (copied it to USB drive, booted from USB drive, formatted the card and copied the USB drive to the card) I conducted some more tests and I was being able to replicate all the issues mentioned above, so either they must be occuring pretty randomly or my install is really messed up. I'm really looking forward to hear about current development/fix status of the tool. And I'd be glad not to hear that I'll have to do a fresh install. Setting up all those little things here and there all over agin would be a nightmare :/
Using Pi as a full-time desktop PC since January 2019

wh7qq
Posts: 1439
Joined: Thu Oct 09, 2014 2:50 am

Re: Raspberry Pi Diagnostics feedback

Sat Mar 07, 2020 1:21 am

I have searched this thread for "diagnostics" and on line with both Google and Duck Duck Go and nowhere has there been any hint of how to get this utility or how to install it. Only one of the posts from a moderator hinted that it might live on Github but there is nothing else in this thread as to how one of the regular members of this forum might obtain, install, try to use or troubleshoot the script. This approach mainly sows confusion.

For those of us who aren't "In The Know", the code is located at https://github.com/tnyim/cpuvinf along with a nice "readme" explaining what it is and how to use it. Remains to be seen if it works out for the rest of us or not but at least some of the rest of us now have a chance.

K4ktus
Posts: 20
Joined: Sun Jun 30, 2019 5:22 pm

Re: Raspberry Pi Diagnostics feedback

Sat Mar 07, 2020 7:34 am

Using Pi as a full-time desktop PC since January 2019

hippy
Posts: 7378
Joined: Fri Sep 09, 2011 10:34 pm
Location: UK

Re: Raspberry Pi Diagnostics feedback

Sat Mar 07, 2020 10:14 am

wh7qq wrote:
Sat Mar 07, 2020 1:21 am
For those of us who aren't "In The Know", the code is located at https://github.com/tnyim/cpuvinf
I am not convinced that's the code used by the Foundation supplied test code and harness.

K4ktus
Posts: 20
Joined: Sun Jun 30, 2019 5:22 pm

Re: Raspberry Pi Diagnostics feedback

Sat Mar 07, 2020 10:17 am

hippy wrote: I am not convinced that's the code used by the Foundation supplied test code and harness.
Of course it's not
Using Pi as a full-time desktop PC since January 2019

User avatar
bensimmo
Posts: 4552
Joined: Sun Dec 28, 2014 3:02 pm
Location: East Yorkshire

Re: Raspberry Pi Diagnostics feedback

Sat Mar 07, 2020 10:45 am

You can find the program and a way to test USB drives etc in the folder.
It's all command line stuff with a GUI for the desktop.
/usr/share/agnostics
(See the blog for more info, near the bottom.)

Code is not available yet, the GitHub repro has nothing in it we can see.
https://github.com/raspberrypi-ui/agnostics

Remember, most cards would probably fail, especially if you're still using you older carss.
They have picked a fast modern A1 spec as the baseline.
It doesn't mean your card will not work,nor is it about to fail. It just means the Desktop GUI experience is not as good as it could be.

"We are testing against the A1 specification, which requires a sequential write speed of 10MB/s, 500 random write operations per second, and 1500 random read operations per second; we run the test up to three times."

(Just info collected from the blog)

Donnelly
Posts: 90
Joined: Wed Aug 21, 2019 10:54 pm

Re: Raspberry Pi Diagnostics feedback

Sat Mar 07, 2020 10:20 pm

Can the Pi Diagnostics be run from the command line and if so how?
Thanks,
Dave

User avatar
bensimmo
Posts: 4552
Joined: Sun Dec 28, 2014 3:02 pm
Location: East Yorkshire

Re: Raspberry Pi Diagnostics feedback

Sun Mar 08, 2020 7:34 am

I thought I had added it above but I guess missed that part

sdtest.sh is the script you need to run
It is in /usr/share/agnostics folder

To change drive you alter a line in sd_bench.fio

tinker2much
Posts: 174
Joined: Wed Jun 20, 2018 12:38 am

Re: Raspberry Pi Diagnostics feedback

Fri Mar 20, 2020 3:14 am

I just ran the new SD Card speed test on my fleet of Raspberry Pi's - zeros, zero Ws, pi2, pi3b, pi3a+, and a pi4, and got QUITE a range of results, some passing, some failing.

In my tests, the "Sequential Write" and "Random Read IOPS" tests were less variable (2.5x between worst and best) and less below spec, and the "Random Write IOPS" test was the MOST variable (14.5x between worst and best) and the most below spec. I was a little amazed by just how bad some of those were on the write test (like 62 or 165 or 209, when the target is 500).

The best kinds for me were Samsung Evo+ 32GB, the Sandisk Ultra Plus 16 or 32, and the Microcenter house brand 16GB cards. I have some of those as spares, so I presume that I could just copy the images over from the bad cards, and get definite improvement right away. (Except for one marginal Ultra Plus, the worst scores came from other brands and models.)

Or would it be worth trying to hard format (write over with zeros) the worst cards, and then copy the images back using SD card copier? I saw some mentions of old cards getting in a state where such formatting could make a big difference (although I also sensed some skepticism.)

Opinions? What's the better thing for me to do?

User avatar
bensimmo
Posts: 4552
Joined: Sun Dec 28, 2014 3:02 pm
Location: East Yorkshire

Re: Raspberry Pi Diagnostics feedback

Fri Mar 20, 2020 8:33 am

It's just the type of card.
The spec 'A1' they are aiming for is new* and hard to achieve. Most cards would fail.
The old Class10 generally would.

The EVO+ was one of the cards that kick this of and became the default go to card, so much they are now on the ISS.
They are not A1 rated, but perform at the level.

Your cards are not bad, nor are they failing. It is a horrible term to use, it should be 'not as good as it could be' and then possible for really slow cards 'not suitable for the Desktop'
It just means they are not as quick as others available that make using the Desktop nicer.

It would be a it like using an high end NVME SSD as a pass fail criteria and then telling you all your other NVME and certainly SATA drives are failing.

No amount of formatting will make them that much faster. I've not actually found where that evidence comes from w.r.t uSD cards.

hippy
Posts: 7378
Joined: Fri Sep 09, 2011 10:34 pm
Location: UK

Re: Raspberry Pi Diagnostics feedback

Fri Mar 20, 2020 11:15 am

bensimmo wrote:
Fri Mar 20, 2020 8:33 am
Your cards are not bad, nor are they failing. It is a horrible term to use, it should be 'not as good as it could be' and then possible for really slow cards 'not suitable for the Desktop'
It just means they are not as quick as others available that make using the Desktop nicer.
A percentage result may be better rather than an absolute pass/fail. Then the user can make an informed decision as to whether to upgrade or stick with what they have.

As it stands the user cannot tell how lacking what they are using is. It may be 95% good enough which the majority would likely consider acceptable to keep on using, with not enough gain to be had from upgrading to make that worthwhile.

It perhaps shouldn't be an "Is your card good enough for desktop use?" test but an "Is it worth upgrading your card?" test.

User avatar
davidcoton
Posts: 4883
Joined: Mon Sep 01, 2014 2:37 pm
Location: Cambridge, UK
Contact: Website

Re: Raspberry Pi Diagnostics feedback

Fri Mar 20, 2020 11:22 am

hippy wrote:
Fri Mar 20, 2020 11:15 am
It perhaps shouldn't be an "Is your card good enough for desktop use?" test but an "Is it worth upgrading your card?" test.
Maybe the test should indicate what standard the card met (Class 4,6,10? A1?) rather than pass or fail. Most users are capable of comparing that with what is written on the card.
Signature retired

tinker2much
Posts: 174
Joined: Wed Jun 20, 2018 12:38 am

Re: Raspberry Pi Diagnostics feedback

Fri Mar 20, 2020 1:15 pm

I understand that this test's pass and fail are WRT an arbitrary standard that may or may not matter for one's own application..

It HAS shown me that some of my cards are MUCH worse than others, and, since I have better cards already on hand, I will try improving performance (whether that's critical or not) by replacing those few cards.

The Microcenter cards in fact consistently (though only modestly) under perform on the sequential test, while consistently over performing on the random tests, and I think that makes them a perfectly good choice for general Pi use, especially since they're also consistently cheap - $3-$4 apiece.

I do agree that the test could somehow be more nuanced than "PASS" and "FAIL".

User avatar
bensimmo
Posts: 4552
Joined: Sun Dec 28, 2014 3:02 pm
Location: East Yorkshire

Re: Raspberry Pi Diagnostics feedback

Fri Mar 20, 2020 2:09 pm

The reason I don't like the pass/fail wording is because people think their cards are faulty (when they are not, this is not a fake card tester etc) and then try to make make them better, which will not happen as they have only just bought a cheap generic (may even a Kingston ) card.

Simon (I assume it's Simon who made it ?) need to note that users will not actually know what is is telling them. There is no background info (even then they may not understand it). They are just give speed test pass/fail from what I remember when running it a few time.
I don't even think it mentioned 'for desktop use'

For terminal use or data logging or simple programming, like the cards use to be they are perfectly fine.


Anyway enough of that, I'm sure they expand and develop it.
I'd like to see.
-firmware checks, are they up to date and then information on what to do (or an automated script at the click of a button)
-check for WiFi locks as that often crops up.
-What Pi version etc this is. (many don't seem to know)


I'd love to know what they have planned for it, what else do people think is faulty ?

spl23
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 402
Joined: Fri Dec 26, 2014 11:02 am

Re: Raspberry Pi Diagnostics feedback

Fri Mar 20, 2020 2:25 pm

wh7qq wrote:
Sat Mar 07, 2020 1:21 am
For those of us who aren't "In The Know", the code is located at https://github.com/tnyim/cpuvinf along with a nice "readme" explaining what it is and how to use it. Remains to be seen if it works out for the rest of us or not but at least some of the rest of us now have a chance.
I have no idea whatsoever what the code in that repo is, but it is nothing to do with us and is definitely not the agnostics source, which is located at https://github.com/raspberrypi-ui/agnostics

Return to “General discussion”