intel / edge-software-provisioner Goto Github PK

I'd love to add a URL to a bash script that gets called at the end of the provisioning (of any Linux based profile).

I've not found mention of one, but I thought I'd check before i go deeply in the wrong direction :)

Ideally, this script would be generated by ESP based based on running an ESP side customisation hook

yes, I'm pondering how to make a pre-FDO magic deployment and onboarding network...

Hi, Maintainers,

How can we shutdown ESP services safely after I finished the node provisioning?

There's no script to tear down the services.

I tried to run "docker-compose down", but after the service containers are down, I failed to delete the ESP folder and got the following "Device busy" issue:

rm: cannot remove 'esp/template/nginx/index.html': Device or resource busy
rm: cannot remove 'esp/data/srv/tftp': Device or resource busy
rm: cannot remove 'esp/data/usr/share/nginx/html/tftp': Device or resource busy

I have to manually "sudo umount" those locations, then I can clean up the folders.

Please help to check this issue and let me know if anything wrong.

Thanks!

As far as i have seen, the run.sh script starts all the containers listed in the docker-compose.yml file, except the gitea container, which is started by a subscript.
So when i want to stop or deactivate the ESP, i always have to stop the "gitea" container manually.

I would suggest that all running containers should be stopped by "docker-compose down".

Hi ESP Team,

I not appear/view in documentation for ESP FORWARD keyword.
I think is necessary to specify lines for FORWORD and NAT for deploy correctly OS target.

Ref: https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/4/html/security_guide/s1-firewall-ipt-fwd
sysctl -w net.ipv4.ip_forward=1

ESP services buid(build.sh) fails when executed inside a container
While executing build.sh inside a docker container "apk update" inside alpine container fails since proxy parameters are not passed to the alpine container hence http://dl-cdn.alpinelinux.org/alpine/v3.2/main/x86_64/APKINDEX.tar.gz is not accessible

While running the command on Ubuntu20.04 it fails. Seems like a bumpup is necessary as the files dont exist anymore

tail builder.log
Step 9/15 : RUN wget http://mirrors.edge.kernel.org/ubuntu/pool/main/l/linux/linux-image-unsigned-${UBUNTU_RELEASE}_${KERNEL_VERSION}.deb && wget http://mirrors.edge.kb
---> Running in 0b428fd08d80
--2022-04-21 16:38:46-- http://mirrors.edge.kernel.org/ubuntu/pool/main/l/linux/linux-image-unsigned-5.8.0-56-generic_5.8.0-56.63_amd64.deb
Resolving mirrors.edge.kernel.org (mirrors.edge.kernel.org)... 147.75.101.1, 2604:1380:2001:3900::1
Connecting to mirrors.edge.kernel.org (mirrors.edge.kernel.org)|147.75.101.1|:80... connected.
HTTP request sent, awaiting response... 404 Not Found
2022-04-21 16:38:46 ERROR 404: Not Found.

ESP uses port 80[HTTP] and port 443[HTTPS] to host nginx server. During provisioning, http and https calls route to their default port numbers. Is there a way that we can specify in ESP to change these default ports to user defined port numbers ?

If there is a feature to detect the hardware its booting from and flash the alternate hardware (sda/b or nvme ) it would be helpful.
In a scenario when the USB was used for the boot, it started wiping the USB device and flashing the device itself

run.sh has docker-compose up -d core web certbot registry-mirror squid mirror smb

The Readme.md has:

• dnsmasq (provides DHCP and TFTP services)
• nginx
• squid - optional, used for caching http requests
• registry - optional, used for caching Docker images
• Gitea - optional, used for mirroring git repositories

considering that core is both privileged, and uses all the host namespaces, this deserves more exposition.

Hi ESP Team,

I saw on the documentation page that an implementation with Yocto or a Yocto derivative would be possible.
In my case I would like to deploy with AGL (Automotive Grade Linux) on my development target.
I would like an example in this case or if you can create a page for this aspect.

Thanks.

Hi, there
We try to build and run latest version of ESP.
When the target machine is configured as Legacy boot with PXE enabled, it report : NBP is too big to fit in free base memory.
When the target machine is configured as UEFI bios mode with PXE enabled, it report: nothing to boot no such file or directory (http //ipxe.org/xxxx)
So is there any pointer we can debug more ?
I think probably because the DNSMasq is not well configured, but we don't know how to proceed the next step.
The topology of the network is one router with DHCP enabled which can access external network, and 2 machines are connected to the router, 1 is ESP server and the other is the target machine.

Don't run build scripts as root on the host. :)

• esp_dnsmasq_1 container failed to start if host system already have libvirt virtual network (virt-manager) running in ubuntu 20.04.
• this container is only able to start if libvirt virtual network is being turned off

I see mention of running build.sh -S if the profiles are changed, but would like to know what needs doing if other parts of the file are changed.

And similarly, assuming that my build system is separate from my run time system, what non-image things need to be transferred - there's a lot of run time volumes, but its not entirely clear if they're just cache / run time data.

We found an issue where ESP is stuck if there is no match found in the dynamic profile which then needs user intervention to boot it again.

It is P1 - blocker for us to start using ESP for our mass deployment

Hi ESP Team,

I try to use this step: (8/9) Building custom files for ... under Ubuntu 20.04 profile, but not working.
I think is one issue with entrypoint variable in scripts/bulkfileutils.sh.
Because of this, the docker container is no longer turned on.

ESP: branch master.

// profile - cat conf/files.yml
---

build:
  # Identify the container to run the build process in
  - container: ubuntu:20.04
    entrypoint:
    # Interrupter to use for execution
    cmd: /usr/bin/bash -c
    # Script to run from the /build/ folder
    execute: start.sh

// esp - git diff scripts/bulkfileutils.sh
@@ -252,10 +252,10 @@ processBuilds() {
                 local container=${files_config_build__container[j]}
                 validateInput container "${container}" "'container' in files.yml of profile ${profileName} is not valid container name: ${container}"
             fi
-            if [ "${files_config_build__entrypoint[j]}" == '""' ]; then
+            if [ -z "${files_config_build__entrypoint[j]+x}" ]; then
                 local entrypoint=""
             else
-                local entrypoint="--entrypoint=\"${files_config_build__entrypoint[j]}\""
+                local entrypoint="${files_config_build__entrypoint[j]}"
             fi
             if [ -z "${files_config_build__cmd[j]+x}" ]; then
                 local cmd="sh -c"
@@ -301,4 +301,4 @@ processEmbeddeds() {
         printDatedMsg "  ${T_INFO_ICON} No files to embed into uOS"
         logMsg "No files to embed into uOS"
     fi
-}
\ No newline at end of file
+}

Profiles do not allow a "$" in the password.

Hi ESP Team,

I use this profile for install Clear Linux OS.
But i view one issue under deployment.

Manifest.Mom.tar is not downloading in chroot.
I think this is cause not appear file /etc/resolv.conf in chroot.

swupd bundle-add desktop
swupd tool not working. MoM file is not downloading.

Profile:

• git_remote_url: https://github.com/intel/rni-profile-base-clearlinux.git
  profile_branch: desktop
  profile_base_branch: master
  git_username: ""
  git_token: ""
  name: clearlinux-desktop_profile
  custom_git_arguments: --depth=1