alepzi: that's wonderful!

so every jail takes a min of 2.5MB ram. looks like thick jail is 2.2 gb on disk for some reason. either way not terrible

the minimal ram is really nice

2.5mb is low overhead enough to basically make jails free for ram

weird. i added the vnet flag and all of a sudden the jail is now running syslogd and sendmail

hmm

jail_reverse_stop="YES" seems pretty useful

any reason to NOT set it?

jail_reverse_stop is perfectly resonable.

If you have set exec.start="sh /etc/rc"; then by default it will start up all of the normal things that a system starts up.

syslog and sendmail can't start on a non-vnet jail because they are already running in the host and can't bind to the sockets they need to bind to and so fail out.

ah, it's the default freebsd env

understand

But once you start a vnet jail then they have their own network stack and the startup for them works.

But you probably don't want them started.

yai'll customize the jail's rc.conf just like i do in the host now

Set sendmail_enable="NONE" to ensure sendmail does not start.

Dan Langille posted this a while back gist.github.com/dlangille/ce60ac76b69f267a3f1de33495a338fc

Title: Periodic things to turn off in FreeBSD jails · GitHub

Probably you will want to make similar customizations to turn those off in jails.

nice

why does the handbook say for a jail to: zfs create ... /usr/local/jail and zfs create /usr/local/jail/containers, but not zfs create /usr/local/jail/containers/testjail (1 for each jail)?

I am not looking at the Handbook but isn't that just one example of one way to do it? There are many ways to do it. It's flexible.

but i mean isn't it a good idea to create 1 dataset per thick jail?

Sure.

alepzi: that can become a management nightmare, as it will increase your zfs list output. what would you like to gain by making a set of zfs datasets per jail?

if you delete a jail.. you would have to remember to delete the zfs

more isolation, maybe being able to individually snapshot a jail, set a quota for size, stuff like that?

this is a curiosity question: what do you mean by more isolation?

i have never thought of the situation, the way you are describing it and curious

like being able to set a size quota on 1 jail's dataset instead on all jails

disk size

snapshot the data of a single jail rather than snapshotting all jail's data at once

yeah i get that.. so isolation is not segmenting the jail itself but the management of the disk utilization of the jails?

ya

Also, if a thick jail is from a zfs clone of a base thick jail, it takes no space (initially)

and that clone will make it its own dataset right scoobybejesus?

indeed

a cheating way could be to look at the bastille scripts. they are all /bin/sh. bastille does two datasets, nested. seems excessive. i wonder how many folks use a modified version of bastille github.com/BastilleBSD/bastille/blo…/usr/local/share/bastille/create.sh

Title: bastille/usr/local/share/bastille/create.sh at master · BastilleBSD/bastille · GitHub

lines 288 and 291

ah, i misspoke. well, at least in bastille, they do zfs create for new datasets, and then do a zfs send | zfs receive after taking a temporary snapshot of the base/"release" jail, plus some odds and ends

down at 404

Hello, is there any document describing how to remove components, like debuginfo from an installed environment? I am using binary based install now.

here comes an odd question : is there a way to tell poudriere that it may use a LOT more time to build a package? Because www/qt5-webengine tends to "timeout" and fail.

blastwave: that actually means the build consumed so much ram that your system OOM'd and the OS killed the www/qt5-webengine build. You need to go to /usr/local/etc/poudriere.conf and config it to either use less builders and/or less make jobs per builder.

hi!, in a server with hdd disks, in a ZFS pool 85% full I need to run the "find" command to get a list of the files in many directories containing thousands of files each. I did't measure it, but it feels very slow. How can I start checking for read speed.

Apart from that, is find the right command to do this?

btw, in a first pass I did a find to collect the directory names (I needed only those containing a subdirectory named 2023-"

sorry, "2023-*", so I have a file listing all those subdirectories. Now I can do an "ls" on those or a "find", I'm looking for the fastest way

it seems "ls" is the way to go

anyway it still looks slow, I'm looking forward to speed up zfs reads.

/usr/bin/time -h dd if=/dev/zero of=sometestfile bs=1024 count=10

never mind, they elft

if anyone is a tmux guru, i'm stumped as to why send-keys -M (pass mouse events to panes) only seems to not work on FreeBSD. test case: start tmux with default config, set -g mouse on, then open something like htop and try to click a column header to sort

i've tried enabling the two unselected config options for the port, but no effect

outside of tmux mouse events work fine, and inside tmux mouse events work fine EXCEPT passing through to panes

Hello! I'm using bash as user's default shell in FreeBSD 14.0. I defined some aliases in $HOME/.bashrc, but if I open a new window in tmux the aliases are ignored. If instead I manually run `source ~/.bashrc', then the aliases work correctly. What could it be the reason?

rockyh, does "echo $SHELL" show bash?

for stuff like that .profile or .bash_profile is usually better. i'd read up on the differences between .bashrc and .bash_profile and choose the right one for how you're using it.

martinrame: try fd-find

Find is slow with how it's implememted

markmcb: yes, but it shows `/usr/bin/bash', which is a symlink to the actual bash path, `/usr/local/bin/bash' (I needed to do this due to some other issues)

according to bash(1), the file used for non-interactive shells (as in my case, with tmux) is ~/.bashrc

the alias is correctly active instead if I specify it in $HOME/.profile

sorry, I was not meaning `non-interactive' shells, but shells that are not login shells, as those run by a new tmux window

rockyh: are you sure, the shell tmux starts is a bash? e.g. have you started tmux while SHELL=/usr/bin/bash was set?

rockyh: e.g. what does `ps -fp $$` shown in a fresh window?

ridcully: IIUC yes, tmux is started within a shell where $SHELL is equal to `/usr/bin/bash`

PID TT STAT TIME COMMAND

67339 9 Ss 0:00.01 -bash (bash)

rockyh: is 14 relevant here? as in: did it work before?

I'm not sure, because I was not using bash with 13.1 or similar

so I can not make a comparison

is it some shenanigan with bash being symlinked then? do you see the same problem if you do `SHELL=/usr/local/bin/bash tmux` ?

also is your ~/.bash_profile sourcing the .bashrc?

the alias works if I run tmux as `SHELL=/usr/local/bin/bash tmux`, so yes, it seems related to the symlink

no, my .bash_profile does not source .bashrc

(actually I don't have a .bash_profile at all)

add `set -x` or some echo in your rc-files then to see what is happening. and create a .bash_profile for the sym-link variant just in case

ridcully: putting `set -x' both in $HOME/.profile and $HOME/.bashrc shows that the latter is completely ignored. I also put `source $HOME/.bashrc' in $HOME/.bash_profile (just created). Thanks for your suggestions!

rockyh, Usually in ~/.profile one "protects" the sourcing of .bashrc to only if the shell is bash and not sh since sh also reads .profile. Usually "if [ -n "$BASH_VERSION" ]; then if [ -f "$HOME/.bashrc" ]; then . "$HOME/.bashrc"; fi; fi; Personally I find it easier to use a .bash_profile for all of my bash specific startup. And then leave .profile for sh to use.

Also if one is using an xdm like lightdm or slim or something then use ~/.xsessionrc for the same as that is the equiv profile for xdm sesssions.

rwp: thanks, it's a good alternative!

termbin.com/rxix with that rc.conf why does ping 10.0.0.1 fails pls?

folks, greetings... not so quick question but I wonder if you could help me...

I have executed `zpool import something` from one FreeBSD environment I have.. into other.. and I regret of doing so..

how do I revert this operation?

(e.g.: I'm my laboratory I have two environments, one I boot with "zroot" ZFS pool and other environment I boot with "vnxsystem" pool

... they are not in mirror, or raid mode, I just came to use two distinct indeed)

but the question remains the same: reverting an zpool-import(8) which seems to have messed with my mounting points.

vinnix, Normally you can only import one pool name at a time. That is to say that if both pools are named "zroot" (the default) then the second import will fail. This tells me that the names are different names. Is that true? And in that case the second one will mount on top of the first one. is that true?

Can you pastebin us a "zfs list" to show the current state of things?

If it is on the network then "zfs list | nc termbin.com 9999" is sufficient and then it will produce a URL back that can be shared.

alepzi, Perhaps you could share the output of "ifconfig" as that would show the current state of things there?

I am not familiar with "autobridge_* configurations at all. I would need to read it. Where are you trying to ping from? From the jail? Or from the host?

rwp, indeed itś true, it seems one mounted on top of each other. we do have two distinct complete pool names, one zroot, and the first one I had the "brilliant" idea of changing it to "vnxsystem" during my setup

rwp, so... to paste I will have to ask for patience and wait me to reboot :P if you dont mind.. we could keep this async comm

rwp, but you got it right, two distinct pools, one one each SSD, mounted on top of each other... odd that was allowed to overlap mounting points.

Honestly I think vinnix's easiest solution is to reboot without the extra vnxsystem pool as that will solve the problem. When they return I will suggest: zpool import -R /mnt vnxsystem

rwp: may I ask the netcat command aqui to share the paste you are looking for?

vinnix_, If you rebooted then all will be okay, right? Because that would have unmounted everything? No?

If you need to mount a second root level zpool on a system with an existing root level zpool then use the -R option to mount to a directory.

zpool import -R /mnt vnxsystem

That will set the altroot propert for that import to /mnt so that mounts will happen there rather than on top of the current directories.

Other than rebooting I would have tried "zpool export vnxsystem" and I think that would have unmounted and exported all okay. Except often things become "busy" and busy mounts cannot be umounted and things get messy. Rebooting was the easier fix.

rwp, my mistake here was not to provide the "-R /mnt" gist.github.com/vinnix/2a2e9781455f28cf93e8ece06ad3ac52

Title: gist:2a2e9781455f28cf93e8ece06ad3ac52 · GitHub

vinnix_, Since that was a df and not a zfs list I don't know if you have things figured out or not?

rwp, okay updating here with zfs list, apologize. in my case, I have tried the reboot and it have not fixed.. just in case

When working with zfs one normally needs to see the "zfs list" information rather than "df" information. I know df is the traditional unix command but zfs does more and that additional information is not displayed with df. So one just needs to use zfs list sometimes.

I think "zpool import -R /mnt vnxsystem" should do it. Unless the mount points are "busy" in which case that will fail.

rwp, I have added the `zfs list` to the gist link above if you wanna see it. I got what you mean, indeed more information

I am going to assume that things are "busy" and the umount will fail. Options 1) ps and lsof to find processes holding those mount points busy and killing those to make those mount points not busy. 2) Booting single user mode so that nothing starts and then those will not be busy then 3) setting the altroot property on vnxsystem to /mnt and then once again rebooting so that the altroot will mount to /mnt on the next reboot.

rwp, okay I may trie (re)import again ...? is that what you are saying? I could no problem

No. Importing again would attempt to mount again. That shouldn't work. But if it did it would just create yet another mount layer and not what you want.

man zpool-export

export is the opposite action to import

What's the physical device situation? Can you just shutdown, detach the vnxsystem drive, then boot again? That's the simplest answer.

If the vnxsystem drive is not present then it will not be imported at boot time. But the zroot will be there. So detaching the vnxsystem so that it can't be imported at boot time is the easy way to do that.

If you can't remove the vnxsystem device then of the choices I think of I would boot single user mode and export it from there to unmount it, set the altroot=/mnt property, then reboot normally from there.

Just a by the way because I need to go afk myself for a while but I have done the same mistake myself before. When I did it that's when I learned about the -R /mnt option. So you are not the only one to have made that mistake. I am sure you won't be the last either! But it is not difficult to recover from it.

Good luck!

rwp, got it, sorry my delayed here, trying to think in my next steps while reading.

rwp, I have updated the data with physical information (dmesg, gpart show)

rwp, understood, `zfs export` next steps to try, I read the manual and it was not clear

Delays are okay. You are hanging around to interact! That's the important point. :-)

small nitpick I notice manuall on zfs-import is duplicated

But I will need to drop offline for a bit. I think you know what you need to do to recover. And if you have other problems one of the other folks will be able to help!

manual duplicated? Which part/

Which part?

rwp, I'm a bit afraid of detaching physicals from the pool and not being able to boot it again from the other environment

rwp, my setup is those "dual-boot" from BIOS time

(reading detach manual)

rwp, correct if I'm wrong here, but since itś a "dual-boot" and I want to be able to remove one disk and boot in another machine in case I have two

rwp, detaching from the device may break it more? or I missed the point? is detach/attach associated with the environment I'm or the disk itself?

When I said detach it I meant unplugging the cable from the drive.

There is no man page for it. The two man pages of interest are zpool-import and zpool-export

sorry them eheh, I thought we were talking about the commando

What did you do to import it?

In the beginning everything was okay. But then what steps happened that caused things to not be okay?

That's what we want to reverse and undo.

I look at man "zpool-import" and I don't see any duplication in the man page, where should I be looking? man.freebsd.org/cgi/man.cgi?query=z…import&manpath=FreeBSD+14.0-RELEASE

Title: zpool-import

rwp: so i fixed my desktop yesterday. but i ended up reinstalling brand new

i need to look into those docs and learn more about zfs

It's all a learning experience! :-)

And I must run off as real life is calling. I'll be back in a few hours. Good luck to all!

have a good one rwp

rwp: well, i ended up updating my dotfiles to at least have an environment ready quicker

so i would call that progress. not ideal as all that docs, but its some progress