Call for testing: IIAB pre-installed on ubuntu-mate-22.04-desktop-arm64+raspi #3303
Comments
|
First Issue: If the screensaver engages your user password will not unlock the session, hard power cycle. Disable the power management and screensaver before they cut in. Should somebody boot the stock upstream image please check for this behavior since I never booted a stock version of the ubuntu-mate-22.04-desktop-arm64+raspi.img source image as my build system is automated. This is most likely an upstream issue that was inherited. |
|
Disable the setting here: Looks like an upstream issue and might be resolved when connected to the internet for updates |
|
Taken after providing internet access via tethered cell phone prior to reboot, seems stable with the screensaver turned off for now. http://sprunge.us/NXQi57 |
|
I've guessing it might be a bad microSD card (maybe, in any case overnight it froze [still hanging now] while writing to the card, 32% of the way through). So I need to try again flashing the 3GB image in another way. |
|
Another microSD card did not work. Hopefully that was just extremely bad luck. I'll try a third one. |
|
3rd microSD card worked. Just FYI Ubuntu Mate had problems sync'ing with the screen (1st minute warnings came up on monitor saying so, 2nd minute the screen worked but was flashing at a very high frequence, 3rd minute it all started to work). After I set a username/password, Mate appears to be now installing many things (looks like it might take 20-30min; I have to step away but will come back to it later today). |
|
MATE completed its many updates and rebooted ✔️ Very unfortunately however http://box and http://box.lan do not work (@georgejhunt very nicely made http://box.lan auto-launch when IIAB is installed on RasPiOS with desktop!) Background: this shortcoming is problematic, as teachers/students depend on this to get to know IIAB. Long story short IIAB communities and IIAB documentation rely on this heavily, so the onboarding + learning experience is uniformly explainable + tangible across all devices — including IIAB's that happen to have a graphical desktop. |
|
The P-400 and small RCA hdmi TV worked great for me, what are you using for a monitor? Like a said it's a full pristine setup, at least you don't have the IIAB install to wait through, just rename the builder,yml file to local_vars,yml and reboot as cmdserv needs to have the local_vars,yml file present or the service doesn't start up. The hotspot is disabled to allow client upstream wifi during first boot without potential interruption, given the image was build in a container wifi was not detected nor the alternate firmware installed, just a pass trough admin-console, the wifi hotspot will become active. |
It's an ancient Dell HDMI monitor, no worries that's it's ugly during the 1st two minutes (screen works in the end!) |
There never was any attempt to port that desktop pop-up to Ubuntu/Mint given the many desktops that could be in play, that is apples to oranges whining. Now the cookie cutter is made with a narrow choice of one desktop George could put the same effort into this desktop if he so chooses.
Sorry, I used the 'normal' name for the machine localhost not the hack when I accessed the admin page from the iiab box. Just checked, http://box.lan does currently work, but I'll check right after first login next time I install a fresh image. |
|
Now there is a target desktop is all I'm saying.
Let me know, this was a first cut, beta release
Get off the soapbox, with the 127.0.0.1 hack in /etc/hosts out of the box should of worked. If that didn't do some quick checking in /etc/hosts if it's still there or not, firstboot might of wiped the change, didn't notice as explained above. Minor issue, easy worked around, stop being a naysayer the second cut will be done on the P-400 resulting in iiab-hotspot-on having the entries for detected AP support then hotspot might be available on firstboot but I'm a little hesitant to do that until there is a user created that can ssh and I would like to make the 5G band available during firstboot for updates. For today with a local_vars.yml file in place just run Maybe for bonus points when the terminal opens banner "please run cd /opt/iiab/iiab and sudo ./iiab-network" like other the warnings do and have the terminal pop open when logging into the gui session? |
|
I'm not being dismissive. I'm just saying I've been in dozens of schools where http://box and http://box.lan have to work (otherwise there's simply chaos). |
Many people will open a browser but not a Terminal. But as a short-term / intermediary solution something like the above (mandatory reminder/nudge in Linux terminal) is definitely pointing in the right direction. |
|
On a side note with /etc/iiab/install-flags/iiab-complete running |
Sorry was evaluating every other aspect of the install, so that comes back to checking /etc/hosts prior to running iiab-network which I proceeded to do because that is the right thing to do to update ap0's mac address per installation or there will be mayhem with duplicate mac address in the air waves when more than 1 of the same image are booted close by. This goes for ALL iiab prepared images without exception so that part should already be documented with this 'issue' going away upon reboot. |
Currently a reboot does restore access to http://box and http://box.lan As you probably well know; but just FYI I tried 3 reboots of this Ubuntu Mate [from 3GB image, on RPi 4] simply to verify that. (The ideal would be that both names |
|
That is what the confirmation that I asked for is about, I'll check it on the next burn/boot, but I'm short of spare sdcards atm...
|
|
Yes I used rpi-imager
What confirmation? I'm away most of the rest of the day, but will send if I can. |
|
If /etc/hosts has the needed entries to resolve box box.lan just after firstboot login, I suspect the file was rewritten from what was originally on the image that I just confirmed is present using mount-image.sh |
|
Just FYI:
PS About 4 apt updates were manually applied prior to the 4th reboot. So possibly that's related to the Bluetooth complaint. |
I'm showing 3 packages to update since I lasted updated. no jumpy screen here
I have a nice blueman-applet in the top right corner.
The top one looks like something the installer did, and emacs would be you, the other 2 are you or the system? I used the provided gui tools to update the system. |
|
This is where I ask for iiab-diagnostics |
Likewise, it's there now. Not sure if it was there after the 4th boot.
The top one happened automatically. The other 4 paragraphs (at the bottom of /var/log/apt/history.log) are from my manual commands. Sorry I haven't time today (to explore it more deeply) but iiab-diagnostics are here if it helps: http://sprunge.us/ZNeEG9?en |
Didn't review the above note, but useful anyway.
In light of the above these would have to be 'True' in default_vars:
as you downloaded 'base':
|
Hum, could tie in 'iiab' instead of exiting at iiab-complete offer to run iiab-network (with new network-complete flag as the final step) after suggesting raspi-config or other related actions (connect to upstream wifi) be completed first. Just ship images with iiab-complete and the in the wild provisioning step would be just |
|
I don't understand all the details but Raspberry hotspots should (wherever possible!) start right up (on 1st boot of image, if feasible?) Running ./iiab-network can and should be heavily promoted...on top of that? Ideally in a targeted or even event-driven way...for those most likely to need it on non-Raspberry HW and USB-smells-like-a-network or Ethernet-link-detected scenarios? (Not sure if that's at all realistic in 2022, but it's an idea.) |
|
FWIW: https://ubuntu-mate.org/raspberry-pi/install/ for a preview of the first boot routine
I've found good performance with 4Gb with the P-400 when that machine was my daily driver and I was running 20.10 and I build pre-updated images from that base image for 21.04 & 21.10 using mk-image with examples which worked fine. That is how I upgraded without risking the running sdcard. I just didn't push to share those images, now I am. |
Just FYI RAM usage was right around 800 MB (i.e. 60% free and available, of 2 GB) immediately after logging in. And about 1.2GB RAM usage after Firefox is launched. But at other times it's entirely possible (similar to Ubuntu Desktop 22.04 LTS itself) that 4GB would be far preferable — especially for anybody using this as they full-time / regular computer. |
You are mentioning observations without context (what monitor, resolution, sync rate) and presenting your opinion as facts. Have a better SDcard to try? Yours might be a bit worn out given the amount of writing they see, nothing lasts forever. I'd start with that given the 7 flashes, means the kernel didn't load at all. |
|
side note: Seeing this across different installs:
Seems related to using |
|
The OS boots about 90% of the time, so the 7 flashes after 5+ minutes of screen sync failing is probably not at all common nor worth worrying about. Normally (during most boots) the repeated screen sync fails for about 30sec-to-2 min (before the 7 flashes would also kick in) which may be ugly, but so be it, it then boots. The particularities of which old Dell HDMI screen are not important (presumably brand particularities are buried on the back of the screen somewhere, but I'm going to worry about it). It's a very generic old HDMI screen that works with all other OS's to date, and so be it. I'm just mentioning this in passing, in case others later run into similar. |
|
Trouble with the boot partition being msdos is editing is really easy from windows, but one has to always remember to 'eject' the card before removing it from the reader or risk damaging the filesystem. |
|
Can get the terminal to open when logged in, lets see if I can get a banner to display. |
|
Actually I did one better, have the complete solution to the must run iiab-network post deployment but I'm not sure if this would be the right place to share it. |
|
I used rpi-imager to flash this image. It came up pretty well. I updated
/etc/hosts with "box.lan" not realizing that reboot was all that would be
required.
I didn't check all the iiab services. But admin-console worked, and it all
seems pretty snappy via remote/wired connection.
But the behavior of the desktop as a console, with usb keyboard and mouse
still leaves a lot to be desired. (hardware was rpi4-4GB). I had tried mint
on Ubuntu 6 months ago and I remember thinking that it wasn't ready for
prime time.
Observations:
1. Click on the firefox icon at the top menu. It finishes painting the
initial screen in 26 Seconds.
2. There is keystroke latency in the search bar of 3-6 seconds which
makes it irritating to correct a mistyped word.
3. Overall the desktop seems to work, but it really feels sluggish. I
tried toggling hardware acceleration in firefox. Seemed better with
hardware acceleration turned off.
4. I tried starting over with a Ubuntu image, and putting lxqt on it
(thinking that mate might be the culprit). lxqt (the newer lxde, which
might not use xorg). It was slow too.
5. Then I installed raspi OS desktop for comparison. Firefox painted the
initial screen in 6 seconds (mate was 26). Chromium was a little slower --
8 seconds.
6. In general, it feels like there is internal misconfiguration, maybe
xorg set wrong, where errors don't bubble up to anything understandable,
but with stalls and timeouts.
…On Thu, Jul 14, 2022 at 5:38 AM Jerry Vonau ***@***.***> wrote:
https://support.pishop.ca/article/33-raspberry-pi-act-led-error-patterns
—
Reply to this email directly, view it on GitHub
<#3303 (comment)>, or
unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAOTQHC2XE7UXUWSWPY7KXTVUAC4HANCNFSM53ICGPFQ>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
|
|
I'd be interested to see iiab-diagnostics from the install. |
|
… On Tue, Jul 19, 2022 at 12:34 PM Jerry Vonau ***@***.***> wrote:
I'd be interested to see iiab-diagnostics from the install.
—
Reply to this email directly, view it on GitHub
<#3303 (comment)>, or
unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAOTQHGBGBFAH2GZYKYXPD3VU37NPANCNFSM53ICGPFQ>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
|
As of today 97 packages can be upgraded from the upstream base image. |
|
Revised images uploaded built with iiab at 4307f65 Intended to help extend the shelf life of images there is a small utility named iiab-upgrade included that could update the install to the latest IIAB code base and runs |
|
interesting idea. how are upgrade_roles and upgrade_roles.txt maintained and by whom. Are upgrades cumulative; if I am one image back do I get the roles I need to upgrade? In the example at hand, the pylibs role should have been included. so will require a second upgrade. To take it further why not compute the roles that have changed by a diff on the last commit on iiab against the commit of the image. For example
|
Initially I can, and for the most part this would be adding roles the would require re-installs that are safe to be added to upgrade_roles and the corresponding note in upgrade_roles.txt. Some role's updates originate when the version is bumped in "defaults" in iiab while others come from outside as when one of the apt installed roles that use 'latest' releases a later version. Ideally long term the person with the best situational awareness of all the moving pieces.
Cumulative, and one could just list both roles within upgrade_roles and would be picked up when iiab-configure runs however because pylibs is not optional the new files are installed with no need to explicitly list that role. Now iiab-configure has used by Adam in the past to move from small -> medium -> large while testing with great success.
Results in: With the endless note editing that approach would have misleading results as above, the narrow focus would just involve the role's bump in the version used by the role. Perhaps when there is some relationship between git commits and 'iiab_revision' that will be possible. |
of course I intended to filter so a role only appears once. could further filter to only pick up roles of interest and installed roles. I agree some changes could be cosmetic and so immaterial. In the end there are not many roles that can be upgraded. When creating the list manually how will you handle the following cases:
|
In the end I don't care how you come up with your solution.
I have had this code proposed in the past as #2500 if adopted then this would be moot point, we could of handled updates since that point in time till the latest code. For the record an older branch has working examples from that time frame. I'm not going to backfit the code to handle that given the past reception the code received and the treatment I received, just shortsightedness in play but now it a good idea. It was a great plan then and still is when you have a 1TB disk full of content and 2 lines changed in the code base for the fix you need.
The last revisions applied would be recorded in iiab.env and upgrade_roles carries the full listing, only the delta between the two would be applied. ie iiab.env has revision 5 applied while upgrade_roles lists 9 revisions 6-9 would be applied when run. |
sounds reasonable.
stops the conversation |
|
Just trying to collaborate, in the end I'm just sharing how I intend to support my release of IIAB within the above images going forward is all. Keeping the basic images devoid of options keeps the delta small between image creation and installed time, that is why I chose 'base' and 'small' rather than the kitchen sink. Now with the presets working rather well just let the end user choose what to install and the content that is wanted once written to disk and booted. Having iiab-upgrade auto update itself gives me a small window into the user's environment that is used to update the basic iiab install and can offer new features going forward. Any support related about the images could be opened at mk-image and iiab-upgrade respectively. |
2 New images for evaluation based on ubuntu-mate-22.04-desktop-arm64+raspi.img
base
small
Exact same user feel as the stock ubuntu-mate images, geared for Desktop therefore use of a monitor/keyboard/mouse is highly recommended. Tested on RPi-400 without connecting to the internet yet. Feedback welcome under this issue, don't file new issues, keep it here.
The image shipped with /etc/iiab/local_vars.yml renamed to local_vars.yml.builder to encourage a review before using, can/should be copied local_vars.yml. Also present is the chroot build script /etc/iiab/builder.sh employed by mk-image as (mate-)runme.sh
iiab/iiab-factory#231
The text was updated successfully, but these errors were encountered: