Custom Query (1118 matches)

While reviewing the tomcat installer for OSGeoLive 14, I have noticed that by enabling the tomcat installer, the live iso boot process breaks so I had to disable tomcat until a solution can be found.

​https://github.com/OSGeo/OSGeoLive/commit/25fbe20f5d07978e77eba4887de1386dab29122a

​https://github.com/OSGeo/OSGeoLive/commit/5665abb2818f8801936486454092ec0f0840a956

In order to figure out what the problem is, I managed to get a snapshot during the iso booting process (tomcat-boot-failure.png). The error is:

adduser: The UID 999 is already in use
usermod: user 'user' does not exist
install: invalid user 'user'

The system (casper) fails to create the 'user' account as the live system user, leaving the system hanging.

Digging a bit deeper, I see the following: On build 37 (without tomcat) right after the startup this is how the /etc/passwd file looks like (live_system_users.png):

...
postgres:x:122:131:PostgreSQL administrator,,,:/var/lib/postgresql:/bin/bash
user:x:999:999:Live session user,,,:/home/user:/bin/bash
systemd-coredump:x:998:998:systemd Core Dumper:/:/usr/sbin/nologin

By manually creating the build chroot environment, right after tomcat installation, this is how the /etc/passwd file looks like (tomcat_user_chroot.png):

...
user:x:1000:1000:user,,,:/home/user:/bin/bash
systemd-coredump:x:999:999:systemd Core Dumper:/:/usr/sbin/nologin
tomcat:x:998:998:Apache Tomcat:/:/usr/sbin/nologin

My understanding of the problem is that tomcat user occupies UID 998 and this makes systemd-coredump user to pick up UID 999, leaving no UID available for the live session user (has to be < 1000). Casper then fails to create the user 'user' leaving the iso hanging on boot.

I see some possible solutions:

1. delete tomcat user in setdown script (as we currently do with user ​https://github.com/OSGeo/OSGeoLive/blob/master/bin/inchroot.sh#L252 ) and find a way to create the tomcat user during boot time.
2. patch tomcat debian package to create the tomcat user with another UID.

Any proposals for other solutions?

While remastering the iso, this command

sudo chroot edit mkinitramfs -c lzma -o /initrd 5.4.0-26-generic

throws an error

/usr/share/initramfs-tools/hooks/cryptroot: 64: cannot open /proc/mounts: No such file
cryptsetup: WARNING: Couldn't determine root device
sed: can't read /proc/cmdline: No such file or directory
grep: /proc/swaps: No such file or directory
/usr/share/initramfs-tools/hooks/cryptroot: 64: cannot open /proc/mounts: No such file
grep: /proc/cpuinfo: No such file or directory

resulting into a non bootable inird file.

Searching through the internet I found several bug reports about mkinitramfs version in 20.04:

​https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1835660

​https://bugs.launchpad.net/ubuntu/+source/ubuntu-meta/+bug/1870260

​https://unix.stackexchange.com/questions/591262/ubuntu-20-04-also-having-issues-while-unpacking-initramfs-but-changing-to-gzip

​https://askubuntu.com/questions/1229964/how-to-extract-initrd-from-ubuntu-20-04-live-server-amd64-iso

​https://askubuntu.com/questions/1245458/getting-the-message-0-283078-initramfs-unpacking-failed-decoding-failed-wh

​https://bugs.launchpad.net/ubuntu/+source/initramfs-tools/+bug/1876562

​https://unix.stackexchange.com/questions/511212/my-initrd-has-only-one-directory-after-decompressing-namely-kernel/511224#511224

​https://www.shell-tips.com/linux/what-is-new-in-focal-fossa-ubuntu-20-04-lts/ --> check last section: "5. Improved Boot Time with LZ4 compression in Ubuntu 20.04 LTS"

OSGeoLive 2020 includes R v4, not the default 3x. details here