Is it expected that bectl can't be run from the install media? I'm using the memstick image for 13.1

crb: yes

except it can be used, the way to use it just isn't documented

the install media use case is actually a really good argument for documenting the undocumented

kevans, ok thanks

there's an undocumented -r flag to bectl that you can use to tell it operate on your pool

is there a way to force safe mode in the loader? My system won't boot without safe mode

i.e., if your pool is zroot and BEs are zroot/ROOT/default and zroot/ROOT/whatever, you pass bectl -r zroot/ROOT list

should be an option in the boot menu for it, yeah

kevans, interesting, thank you

kevans, no I mean I want to add it to loader.conf as I don't want to touch the loader menu every time I reboot

safe mode is a collection of knobs rather than a single knob

what's the failure mode when you're not using safe mode?

kevans, the system reboot just before completing reboot, I can reboot again now if you want precise information

yeah, you might record the console output, it'd be good to narrow down which knob(s) are actually helpful rather than applying that massive of a hammer

the knob list is here, btw: cgit.freebsd.org/src/tree/stand/lua/core.lua#n177

Title: core.lua « lua « stand - src - FreeBSD source tree

so if you set each of those values as described in loader.conf, you have the equivalent

kevans, it gets all the way to the boot prompt and then reboots, the boot messages all go by so fast I couldn't log anything

last resort: video it with your phone?

right, sorry, video is what I meant

kevans: ah, ok that I'll have to try

it'd be good to fix it, but it might also give you a good idea of which of these is actually problematic

but adding those loader configs did work

perhaps I'll work toward eliminating them one by one

crb: yeah that seems reasonable

kevans, the only loader option I have now is kern.smp.disabled="1" which is a bit troubling

kevans, oops not, I guess that's not enough, it reboots doing a simple pkg install git

crb: have you considered that it may simply be haunted?

ghost in the shell is pretty underestimated

kevans: it's a new setup so there are lots of possibilities

but it does seem to want kern.smp.disabled="1"

which isn't good

Jails and rc.conf - is there a way to pass along arguments to the service by setting any values in the rc.conf. For instance postgresql, i want to be able to set the port to use.

and also trying to understand if there is a way to set this by adding config to the jail.conf

can you maybe do it via postgresql_flags ?

nimaje: So I thought when I looked at the rc-script. but if add postgresql_flas="-p 5430" I get "waiting for server to start....exec: 5430: not found" on start

flags* sorry

drobban: -p seems wrong "-p path"

--options= seems to be what you want

thanks. looked at the postgres man page.... Im confused =)

seems like you want the pg_ctl manpage instead cgit.freebsd.org/ports/tree/databas…ql12-server/files/postgresql.in#n25

Title: postgresql.in « files « postgresql12-server « databases - ports - FreeBSD ports tree

nimaje: that explains the confusion

drobban: do i understand you correctly, that you're trying to do this from the host's rc.conf or its jail.conf?

at the moment Im just trying to understand the manual on how to do this on the hosts rc.conf. It would pretty sweet if I also could do it in the jail.conf

or something that achieves the same end result

I don't think there's an easy way to achieve this, without creating an rc.conf for the jail itself

ah, you want to do it from the hosts rc.conf and not from the jails, I don't think it will work then

nimaje: yea understood your answer as that.

meena: bummer. it would be sweet to be able to pass in arguments to the host somehow.

you might be able to shoestring some gross things with environment variables together, but it won't be pretty, easy, nor will it feel very solid

meena: aah, think I might be able to accomplish the result im look for with service -j command

the context of it all is to find replacement tools for my linux dev environment. As I used docker to solve some of my development environment Im now trying to use jails instead.

jails with templates also looks promising.

so the problem I try to solve is not to "maintain" services, but instances that I can set up and slaughter whenever needed with predefined packages and services

good morning

hi

Hi.

drobban: you might wanna look into buildah + podman in that case then

meena: at first glance it looks like just another linux container tool?

drobban: it's actively being ported to FreeBSD to work with jails

might still be a little too alpha

nice

meena: who's working on porting podman?

dkeav: dfr, mostly

Title: 268717 – [pf] [ipnat] rdr rules don't work for traffic originating at localhost

Title: GitHub - dfr/ocijail: An OCI runtime using FreeBSD jail

Title: FreeBSD containers with podman and buildah

Title: Branches · dfr/podman · GitHub

is it common for chromium to crash all the time in freebsd? or am I missing some important config

It doesn't crash so often for me.

And I feel I'm running things in a vanilla way, unmodified.

check.. I wonder what the problem is then.

the crashes are common while on sites with lots of javascript loaded

I find running Chromium in Incognito or Firefox in Private mode helps with both

as for performance and crashes

almost like it hits some kind of memory limit

and firefox does a lot better with js-heavy websites in my experience

as long as it's not a google website :^)

yea, run youtube-music... =)

hummmm... login.conf

okey. unlimited. I wonder whats causing it.... annoying.

Build chromium with debugging symbols and use the core file to find out what's causing the crash.

Eta for debug ports?

and debuginfod?

paulf: hrmmmm, maybe, yeah

just having debuginfo for libc libc++ and libstdc++ would already be a big improvement

meena, I seem to have lost the URL, but I distinctly remember a wiki page (I think by Ben Woods, so maybe in their wiki space) for building ports with debug info.

I intended at some point to work making it into shape for inclusion in the PHB but never found the time.

Ah. wiki.freebsd.org/BenWoods/DebuggingPorts

Title: BenWoods/DebuggingPorts - FreeBSD Wiki

Noooo

I don't want to have to build a port just to be be able de debug it

On Fedora I can either just install debuginfo packages

Or even just type "gdb" and gdb will kindly ask me if I want to use debuginfod

Point. But that change is way above my pay grade.

What exactly am I supposed to be putting in network0_device for vm-bhyve? The bridge? A tap device?

how cringey is it when geli has not been warning for zero keys and encrypting with nothing for 4 years?

wwt: Only when you initialize several in a batch, which hopefully isn't super common.

hopefully

but this is a severe bug

bhechinger: network0_device would be your pre-created tap device

Ok, that's what I thought, but I'm still getting the file not found issue when trying to start the VM. :(

i don't know- i always used the bridges vm-bhyve created 🤷‍

Yeah, unfortunately I can't because I need the one vlan bridge up for NTP/NFS and vm-switch seems to run after that.

you can config th vlan and add it to hte bridge

that might work for you

NAME TYPE IFACE ADDRESS PRIVATE MTU VLAN PORTS

sw200 standard vm-sw200 - no - - lagg0.200

then the interface is on the vlan before bhyve starts

i wish i had this problem bugs.freebsd.org/269572

Title: 269572 – MAXCPU too low for contemporary servers

greetings! why iscsi (istgt) caps @ 800MB/s whereas cam doesn't?

ugh, so it seems my issue with multiple password prompts is due to the older bios bootloader instead of efi. but efi doesn't mirror out of the box. think it's worth the hassle to reinstall and use efi instead of bios so i can just have one password prompt?

new server, no data, just tinkering before i commit to the build :P

otakon: hard to say without flamegraphs

The answer to any performance problem almost always involves flamegraphs.

@debdrup i am using xigmanas, I am no expert in *nix things. However this doesn't seem to be xigma's fault, because when using cam the performance is almost close to the theoretical of the 10gb nic, 1.0~1.1GB/s

otakon: it's no surprise to me that CAM is faster

debdrup my "issue" with cam is that besides being difficult to setup, it uses primarilly the first thread of the cpu, ~70% and the other 3 ~30%. The istgt is more balanced

Why is that a problem? Typically you don't want performance-sensitive workloads moving around on threads, since you'll be losing cache coherence with L1 and L2 (as these are local to each core).

debdrup Is this what happens with istgt?

otakon: I can't answer that without flamegraphs.

debdrup can you tell me how to draw them?

use red and orange crayons

ha

otakon: you don't draw them, you use instrumentation like dtrace

debdrup my bad, thats what I meant

Basically, a flamegraph tells you what part of a codebase, over a duration of time, is spent actively being processed.

Title: FreeBSD Flame Graphs

freshports.org/benchmarks/flamegraph should have all you'll need, if memory serves.

Title: FreshPorts -- benchmarks/flamegraph: Stack trace visualizer

debdrup I get it. But even if something is taking much cpu time, when using istgt I still have plenty of cpu usage left

otakon: flamegraphs will also tell you if it's waiting on I/O

debdrup you are correct. I will try to make it happen (get you the graphs). In the meantime, is it som sort of misconfiguration of the istgt parameters?

otakon: I have no idea what your configuration is, nor do I know what defaults XigmaNAS uses.

XigmaNAS folks ought to be the ones helping you with this.

debdrup unfortunatelly I din't have success, thats why I post here

it seems to be issue with the istgt (of cource I might be wrong)

debdrup shall I pos the parameters?

otakon: if you want, but please use a pastebin. Also, don't expect me to be able to help you, so please don't feel a need to keep hilightingm e.

here is my istgt settings pastebin.com/dl/yY1Sifa2

otakon: is that an InfiniBand card?

also potentially useful: ibv_devinfo -l ; and ibv_devinfo -v <device>