mikishapiro / armbuild Goto Github PK

As an FYI - i was getting an error from the vgrename task in partscan. The error was /dev/fedoraarm already exists. I commented out that task and the playbook ran to completion. A few changes to the raspberry URLs also were needed.

thanks for making this available. I created a fork to experiment with some changes.

Best

Hello,

Thank you for this playbook, this is exactly what I was looking for and the writeups are excellent. After encountering a little issue with the lvm setup of the playbook (#10) it went well and I was able to boot the resulting image.

However I've used the latest links for Fedora 33 and RaspberryOS, and after booting, updating fedora (excluding kernel* linux-firmware*) and installing the workstation group to get GNOME, the wifi/bluetooth is not working.

In my understanding, the playbook installs all RaspberryOS drivers to the Fedora image, so they should have been installed (wifi with plain RaspberryOS CentOS 8 rpi4 (see below updates) works).

I did not notice other upgrade packages besides kernel and firmware that could modify the custom files, am I missing something ?

EDIT: tried reflashing and not updating fedora to see if wifi was working before the update, no luck:

$ nmcli
wlan0: unavailable
        "Broadcom BCM43438 combo and Bluetooth Low Energy"
        wifi (brcmfmac), 92:51:4A:0A:91:65, hw, mtu 1500

EDIT2: tried reflashing with the original images for Fedora 32 and Raspbian specified in this repo's master branch in inventory, got the same result

EDIT3: tried flashing plain latest Raspberry OS to double-check it worked, and now I notice the same behavior: it does NOT work, the wlan0 device is also unavailable. I'm pretty sure I got WiFi working on this pi previously, I think it was with centos8, I will try again with it and keep updated.

EDIT4: i can now confirm that using CentOS 8 rpi4 from this image from this thread has wifi working.

wlan0: connected to <MY_SSID>
        "Broadcom Wi-Fi"
        wifi (brcmfmac), DC:A6:32:XX:XX:XX, hw, mtu 1500
        inet4 10.0.0.137/24
        route4 0.0.0.0/0
        route4 10.0.0.137/32
        route4 10.0.0.0/24
        inet6 fe80::88b3:25e3:xxxx:xxxx/64
        route6 fe80::/64
        route6 ff00::/8
        route6 fe80::88b3:25e3:xxxx:xxxx/128

Thank you

hey, new to ansible so I'm not sure if this is just user error. I made some changes to the script, set the target device and used the same machine for builder1/flasher1 and changed the script to run on a local connection.

everything's working up until task "Remove partition 2 on {{ targetdevice }} on {{ targetmachine }}" on line 40 in flash.yml. the playbook outputs this error message

fatal: [localhost]: FAILED! => {
    "changed": false,
    "err": "Error: Can't have a partition outside the disk!\n",
    "invocation": {
        "module_args": {
            "align": "optimal",
            "device": "/dev/sdc",
            "flags": null,
            "label": "msdos",
            "name": null,
            "number": 2,
            "part_end": "100%",
            "part_start": "0%",
            "part_type": "primary",
            "state": "absent",
            "unit": "KiB"
        }
    },
    "msg": "Error while getting device information with parted script: '/usr/sbin/parted -s -m /dev/sdc -- unit 'KiB' print'",
    "out": "BYT;\n/dev/sdc:266240kiB:file:512:512:unknown::;\n",
    "rc": 1
}

I'm guessing that this script was wrote with the assumption that the target device was new/empty, and after righting the boot and root partitions to the beginning of the device, partition 2 should have been the remainder of the device, but the previous task of writing these two partitions executed successfully. so I'm unsure what the issue is.

If I need to provide more information, please let me know

Hi,

While running the playbook I encountered an issue with the vgrename command, but it was a different issue than #8 .

When playing this action, the UUID from previous action was empty leading to a wrong vgrename use. I investigated why the previous action would not return the UUID correctly and found that a line had been programmatically inserted in /etc/lvm/lvm.conf (the line was just before the closing bracket and unindented which helped noticing it, I found the lvm filter lead by specifically probing /dev/mapper/loop0p3 with vgscan which said 'device excluded by filter'):

filter = ["a|^/dev/disk/by-id/lvm-pv-uuid-ALZV7M-mhBn-GPkQ-HdRz-QHQK-tzb8-sKNHbZ$|", "r|.*|"]

and this UUID happens to be the one of the lvm partition in the image, which is why it was not recognized by lvm after exposing them with kpartx.

After removing that line from the config file, the playbook went up to the end without issues. It was run on a fresh install of centos with all 3 roles being localhost and the raw image file was downloaded by the playbook itself.

Any idea how that insertion happened ?

While debugging #10 I had to reboot my host (of the 3 roles) between runs to make sure the state of devices was clean at start.

I'm brand new to ansible (wanted to play with it for a long time so thanks for the intro) and I've noticed that there is a rescue functionality that could be used to rollback any changes to devices in case of failure.

It could be used to kpartx -d the raw image's partitions and probably other stuff, I'm still learning on all the actions the playbook is doing.

With GNOME, when reinserting the flashed media after a successful execution, the medium gets mounted. It made the playbook failed on my end with "device busy". Ejecting it with GNOME file manager resulted in the playbook failing with "device not accessible".

Could the playbook handle it by unmounting the medium correctly before flashing ? Currently when inserting the drive i'm just formatting it so there's nothing to mount before flashing.