I only started using FreeBSD last year but it was crazy to upgrade to 15 from 14 and be told that pkgbase, while experimental would be a much more unified solution. Come the 15.1 upgrade and you have to run like 5 commands compared to just the old FreeBSD-base fetch command.
Still it’s great to see this work and the desktop-installer script streamline a lot of the experience. I like FreeBSD and have had an otherwise fantastic experience learning and using it.
I've been away for several years as a regular user (only test installations in VMs the last 20 years and no lifecycle management).. but now when I installed 15 and embraced the pgkbase as its the way forward, just to experience a overly complicated upgrade procedure.
I really regret that FreeBSD project did not used dedicated `pkgbase(8)` tool for FreeBSD Base System and leave `pkg(8)` just like it was for 3rd party packages from FreeBSD Ports. That single decision alone will solve many problems and misunderstandings. I generally assumed that people behind PKGBASE concept had everything 'think-out' and they only needed to implement it ... I believe my assumption was wrong and even Colin proved to me when he replied in the FreeBSD Mailing Lists that it was forgotten to find a way to do 'minor' FreeBSD upgrades like from 15.0 to 15.1 in a PKGBASE world ... and I believe another such area is 'tooling' behind PKGBASE upgrade process between releases.
It also reminded me of why Docker become so popular while being less secure and less good solution then FreeBSD Jails - it was all about TOOLING. FreeBSD Jails - and especially VNET ones are really great - but there is just 'limited' `jls(8)` that does not show VNET IPs or poweroff off Jails and `jail(8)` command and `jail.conf(5)` config. You do 95% of the stuff in `vi(1)` in the Jail configs ... the tools are really limited. Even check Solaris/Illumos Zones with its `zonecfg(1M)` and `zoneadm(1M)` tools. Same problem is here now with PKGBASE 'tooling' for upgrading process between both minor and major FreeBSD versions ... and its not better for 3rd party packages. If You want to 'just' upgrade the 3rd party packages - this is the command if You use PKGBASE system:
... and its often that a lot of these tools are available - they are just ignored and not used. Take my `jmore(8)` Jails listing and manager tool. Its different then BastilleBSD or iocage that it does not need any organized way of doing Jails - its fully compatible with 'plain' Jails configured in `jail.conf(5)` file - it is BSD licensed ... but ignored.
# jls
JID IP Address Hostname Path
# jmore -h
usage:
jmore list jail(8) containers
jmore -a list jail(8) containers with all IP(s)
jmore -d INTERVAL run 'jmore' every INTERVAL seconds like top(1)
jmore -h show help
jmore --help show help
jmore help show help
jmore version show version
manage:
jmore <JAILNAME> start
jmore <JAILNAME> restart
jmore <JAILNAME> stop
jmore <JAILNAME> status
jmore <JAILNAME> console
jmore <JAILNAME> shell
jmore <JAILNAME> jexec
shorts:
jmore <JAILNAME> u UP ------> alias for start
jmore <JAILNAME> d DOWN ----> alias for stop
jmore <JAILNAME> r RESTART -> alias for restart
jmore <JAILNAME> c CONSOLE -> alias for console|shell|jexec
jmore <JAILNAME> s STATUS --> alias for status
# jmore
JAIL JID CPU/% RAM/mb TYPE VER DIR IFACE IP(s)
classic - 0.0 0.0 std 13.2-R /jail/classic em0 10.0.0.199
ctld-two - 0.0 0.0 vnet 13.2-R /jail/ctld-two ${if}b -
ctld - 0.0 0.0 vnet 13.2-R /jail/ctld ${if}b -
fbsdjail - 0.0 0.0 std 13.1-R /jail/fbsdjail wlan0 10.0.0.43
iscsi - 0.0 0.0 vnet 13.2-R /jail/iscsi ${if}b -
joudriere - 0.0 0.0 vnet 14.3-R /jail/joudriere ${if}b -
kerberos - 0.0 0.0 vnet - [GONE]/jail/kerberos ${if}b -
minecraft - 0.0 0.0 std 14.2-R /jail/minecraft em0 10.0.0.210
minio - 0.0 0.0 std 14.0-R /jail/minio em0 10.0.0.133
nfsd - 0.0 0.0 vnet 14.1-R /jail/nfsd ${if}b -
other - 0.0 0.0 std 14.1-R /jail/other em0 10.0.0.199
pkgbase - 0.0 0.0 vnet 14.3-R /jail/pkgbase ${if}b -
sambajail - 0.0 0.0 vnet 14.1-R /jail/sambajail ${if}b -
unfs3 - 0.0 0.0 vnet 14.1-R /jail/unfs3 ${if}b -
# jmore classic u
Starting jails: classic.
# jmore ctld u
Starting jails: ctld.
# jmore ctld c
ctld # exit
# jmore
JAIL JID CPU/% RAM/mb TYPE VER DIR IFACE IP(s)
classic 1 0.1 18.0 std 13.2-R /jail/classic em0 10.0.0.199
ctld-two - 0.0 0.0 vnet 13.2-R /jail/ctld-two ${if}b -
ctld 2 1.4 18.0 vnet 13.2-R /jail/ctld epair110b 10.1.1.110
fbsdjail - 0.0 0.0 std 13.1-R /jail/fbsdjail wlan0 10.0.0.43
iscsi - 0.0 0.0 vnet 13.2-R /jail/iscsi ${if}b -
joudriere - 0.0 0.0 vnet 14.3-R /jail/joudriere ${if}b -
kerberos - 0.0 0.0 vnet - [GONE]/jail/kerberos ${if}b -
minecraft - 0.0 0.0 std 14.2-R /jail/minecraft em0 10.0.0.210
minio - 0.0 0.0 std 14.0-R /jail/minio em0 10.0.0.133
nfsd - 0.0 0.0 vnet 14.1-R /jail/nfsd ${if}b -
other - 0.0 0.0 std 14.1-R /jail/other em0 10.0.0.199
pkgbase - 0.0 0.0 vnet 14.3-R /jail/pkgbase ${if}b -
sambajail - 0.0 0.0 vnet 14.1-R /jail/sambajail ${if}b -
unfs3 - 0.0 0.0 vnet 14.1-R /jail/unfs3 ${if}b -
# jls
JID IP Address Hostname Path
1 10.0.0.199 classic /jail/classic
2 ctld /jail/ctld
This looks very nice! The one thing I think would be killer is if it could be integrated with boot environments, so you wind up with something like beinstall where a new BE is created and the update is done in the new BE.
Well, at least I trust the OpenBSD devs not to replace sysupgrade/syspatch with some overcomplicated new hotness, since they only ever make admin tools simpler.
FreeBSD project used freebsd-update(8) for how long? More then a decade I assume. It was introduced in 2009 - and as freebsd-update(8) will be supported for entire 15.x line - the FreeBSD 15.5-RELEASE will be released in 2028 with 2 years of support - that means 2030 - its more then 2 decades of 21 years span.
On the other hand OpenBSD syspatch(8) was introduced in 2017 and sysupgrade(8) in 2019 so they both are younger then a decade.
It would be better to have this discussion in 2+ decades of usage of these OpenBSD equivalent tools.
While I admire OpenBSD team for all the things that are being created 'around' the OS - like:
VMM, abandon all hope for chroot and extended chroots.
>--;
>WINE
Literally in -current
>Chromium and for Signal pick another client
>Real SMP
Unlocking more and more locks
>Poudriere
You might find some tool in ports, no pun intended
On GeFS, in ran fast enough even on my crappy n270 netbook with 1G of RAM and a rotational HD... under 9front. BTRFS would agonize there. ZFS would be hardcore BSDM.
I already mentioned GEFS in my comment ... but its 'just' filesystem - I did not heard about GEFS Boot Environments or anything similar.
> VMM, abandon all hope for chroot and extended chroots.
One can say the same about OpenBSD - abandon OpenBSD and just use FreeBSD.
> Literally in -current
Its a fresh thing - we will see if it sticks.
> Chromium and for Signal pick another client
I can not pick another client as I use Signal Desktop for my daily work - its like saying to my entire team 'fuck you use something else' ... and its NOT happening - I will just use FreeBSD that works for me instead of OpenBSD that does not.
> Unlocking more and more locks
It will take another at least 10 years to be on FreeBSD level with that. FreeBSD already has SMP PF ...
> You might find some tool in ports, no pun intended
That I believe that it can be possible to find something 'similar'.
...
-e errexit
Exit immediately if any untested command fails in non-interactive
mode...
-u nounset
Write a message to standard error when attempting to expand a
variable...
Both `set -eu` and `shellcheck` tools are overrated.
I have used `shellcheck` against many of my scripts and for 'well think and designed' scripts its like a 'false positive generator' tool that only waste my time.
I mean ... POSIX /bin/sh is a language like any other - like Python - like Perl ...
I have been writing scripts since 20+ years in POSIX /bin/sh tools and thinking and knowing what and how a thing can fail always wins.
... and as FreeBSD has ZFS Boot Environments (and even UFS Boot Environments) - no matter what fails and where - You are always protected from failures.
The `set -eu` is treated like some magic 'Save Game' thing - that after You set it in a script - You are always protected from everything ... You are not.
A command that will wipe Your entire / filesystem can succeed without an issue and You will see zero errors.
I only started using FreeBSD last year but it was crazy to upgrade to 15 from 14 and be told that pkgbase, while experimental would be a much more unified solution. Come the 15.1 upgrade and you have to run like 5 commands compared to just the old FreeBSD-base fetch command.
Still it’s great to see this work and the desktop-installer script streamline a lot of the experience. I like FreeBSD and have had an otherwise fantastic experience learning and using it.
I've been away for several years as a regular user (only test installations in VMs the last 20 years and no lifecycle management).. but now when I installed 15 and embraced the pgkbase as its the way forward, just to experience a overly complicated upgrade procedure.
This is a great step!
I really regret that FreeBSD project did not used dedicated `pkgbase(8)` tool for FreeBSD Base System and leave `pkg(8)` just like it was for 3rd party packages from FreeBSD Ports. That single decision alone will solve many problems and misunderstandings. I generally assumed that people behind PKGBASE concept had everything 'think-out' and they only needed to implement it ... I believe my assumption was wrong and even Colin proved to me when he replied in the FreeBSD Mailing Lists that it was forgotten to find a way to do 'minor' FreeBSD upgrades like from 15.0 to 15.1 in a PKGBASE world ... and I believe another such area is 'tooling' behind PKGBASE upgrade process between releases.
It also reminded me of why Docker become so popular while being less secure and less good solution then FreeBSD Jails - it was all about TOOLING. FreeBSD Jails - and especially VNET ones are really great - but there is just 'limited' `jls(8)` that does not show VNET IPs or poweroff off Jails and `jail(8)` command and `jail.conf(5)` config. You do 95% of the stuff in `vi(1)` in the Jail configs ... the tools are really limited. Even check Solaris/Illumos Zones with its `zonecfg(1M)` and `zoneadm(1M)` tools. Same problem is here now with PKGBASE 'tooling' for upgrading process between both minor and major FreeBSD versions ... and its not better for 3rd party packages. If You want to 'just' upgrade the 3rd party packages - this is the command if You use PKGBASE system:
... and its often that a lot of these tools are available - they are just ignored and not used. Take my `jmore(8)` Jails listing and manager tool. Its different then BastilleBSD or iocage that it does not need any organized way of doing Jails - its fully compatible with 'plain' Jails configured in `jail.conf(5)` file - it is BSD licensed ... but ignored.
... and `pkgbase(8)` will also be ignored.
Regards,
vermaden
This looks very nice! The one thing I think would be killer is if it could be integrated with boot environments, so you wind up with something like beinstall where a new BE is created and the update is done in the new BE.
Hi and thank You,
one of the steps is already creation of new ZFS Boot Environment before the upgrade:
I did not created upgrade inside new ZFS BE as I wanted to 'follow' official FreeBSD project instructions - but it cal also be done.
Regards,
vermaden
Upgrading inside a new ZFS BE, as I've seen you comment on, seems like a superior way of upgrading to me.
That may be a 'TODO' for some of the next releases - I will keep that in mind.
This tool looks great and a much better UX for upgrading.
Thank You, trying as always :)
Well, at least I trust the OpenBSD devs not to replace sysupgrade/syspatch with some overcomplicated new hotness, since they only ever make admin tools simpler.
FreeBSD project used freebsd-update(8) for how long? More then a decade I assume. It was introduced in 2009 - and as freebsd-update(8) will be supported for entire 15.x line - the FreeBSD 15.5-RELEASE will be released in 2028 with 2 years of support - that means 2030 - its more then 2 decades of 21 years span.
On the other hand OpenBSD syspatch(8) was introduced in 2017 and sysupgrade(8) in 2019 so they both are younger then a decade.
It would be better to have this discussion in 2+ decades of usage of these OpenBSD equivalent tools.
While I admire OpenBSD team for all the things that are being created 'around' the OS - like:
- OpenSSH
- OpenRSYNC
- doas(1)
- LibreSSL
- OpenNTPd
- OpenSMTPd
- ... and many more listed here: https://openbsd.org/innovations.html
The OpenBSD is just 'too limited' for my needs.
I mean ... I need these for daily operation:
- ZFS/OpenZFS with ZFS Boot Environments
- Jails with VNET
- Hypervisor for ${ANYOS} which Bhyve provides
- WINE
- Linux Compat (for Signal Desktop and Google Chrome for a start)
- 'real' SMP
- Poudriere (or equivalent)
- ... I probably forgot something that is obvious in FreeBSD 'world'
... and OpenBSD 'by decision' just do not provide them.
But I liked the recent news about GEFS (Good Enough File System) - it does not have to be ZFS - GEFS sounds also 'OK' for most needs.
Hope that helps.
Regards,
vermaden
What's wrong with Mozilla Firefox?
Where did I mention Firefox in my reply?
> ZFS
GeFS is on its way.
>Jails
VMM, abandon all hope for chroot and extended chroots.
>--;
>WINE
Literally in -current
>Chromium and for Signal pick another client
>Real SMP
Unlocking more and more locks
>Poudriere
You might find some tool in ports, no pun intended
On GeFS, in ran fast enough even on my crappy n270 netbook with 1G of RAM and a rotational HD... under 9front. BTRFS would agonize there. ZFS would be hardcore BSDM.
> GeFS is on its way.
I already mentioned GEFS in my comment ... but its 'just' filesystem - I did not heard about GEFS Boot Environments or anything similar.
> VMM, abandon all hope for chroot and extended chroots.
One can say the same about OpenBSD - abandon OpenBSD and just use FreeBSD.
> Literally in -current
Its a fresh thing - we will see if it sticks.
> Chromium and for Signal pick another client
I can not pick another client as I use Signal Desktop for my daily work - its like saying to my entire team 'fuck you use something else' ... and its NOT happening - I will just use FreeBSD that works for me instead of OpenBSD that does not.
> Unlocking more and more locks
It will take another at least 10 years to be on FreeBSD level with that. FreeBSD already has SMP PF ...
> You might find some tool in ports, no pun intended
That I believe that it can be possible to find something 'similar'.
Flare Messenger speaks Signal.
An administration sh script without `set -eu` in 2026? I also suggest shellcheck linting.
I’m pretty sure set -eu is for bash; the script is written in bourne.
set -eu is POSIX sh. You're probably thinking about set -o pipefail (which is POSIX 2024 but not supported everywhere yet).
On FreeBSD 15.1:
man sh
Both `set -eu` and `shellcheck` tools are overrated.
I have used `shellcheck` against many of my scripts and for 'well think and designed' scripts its like a 'false positive generator' tool that only waste my time.
I mean ... POSIX /bin/sh is a language like any other - like Python - like Perl ...
I have been writing scripts since 20+ years in POSIX /bin/sh tools and thinking and knowing what and how a thing can fail always wins.
... and as FreeBSD has ZFS Boot Environments (and even UFS Boot Environments) - no matter what fails and where - You are always protected from failures.
The `set -eu` is treated like some magic 'Save Game' thing - that after You set it in a script - You are always protected from everything ... You are not.
A command that will wipe Your entire / filesystem can succeed without an issue and You will see zero errors.
Regards,
vermaden