hello, just updated to freebsd 13.3 from 13.2. the module radeonkms seems to lead to a kernel panic. so, what now? build kernel from scratch?

how do you backup the working kernel as well

I don't want to lose it

nmz, when similar problems happened in the past, the usual cause was a kernel ABI change that sneaked in. If that's also the case here, installing radeonkms from ports instead of packages, or reinstalling it if already installed from ports, should help. You shouldn't need to rebuild the kernel itself.

nmz, I have the same problem here in that I use the radeonkms driver from ports. There are two choices. Compile your own radeonkms against the new 13.3R kernel. Return to 13.2R which is still in support and has a new update recently. I returned to 13.2R for the moment since that was easy.

The easiest thing is to reboot and select the previous Boot Environment. That will return to the previous 13.2-RELEASE from before the upgrade and everything will be good to go again.

The root cause of the problem is that ports are compiled against the oldest supported system version. Which at this time for quartely is 13.2R. That will expire in about another 2 months when 13.2R hits EOL. Then ports will be compiled against 13.3R and so in about two months one could upgrade and the radeonkms in ports should match and work then.

Also let me note again that 13.2R is still in support and a new 13.2-RELEASE-p11 just posted recently. It's not dead yet. No real pressing need to push on at this time.

Or you could push forward to 14.0-RELEASE too. I myself always wait for the .1 release. I am just rather risk adverse. But forward to 14 would be another way to go too.

V_PauAmma_V: make deinstall and make reinstall does nothing, it seems to just fetch the binary

I think its because I never installed /usr/src

rwp: thanks, I just freebsd-update rollback, I'll just wait for the .1 release

any recommendations for youtube channels that have really good vids on zfs?

mostly like the basics, snapshots, clones, copy, etc etc

pools, datasets

Title: OpenZFS

none of those look like basic overviews

not the ones on the main page, but if you click into learning more about zfs

you would find openzfs.org/wiki/Newcomers

Title: Newcomers - OpenZFS

oooo, ty!

zfs is amazing tech. it's the only FS i'll ever run

specifically, openzfs.org/wiki/Newcomers#Where_can_I_learn_more_about_OpenZFS?

Title: Newcomers - OpenZFS

watching it now

any ideas what would trigger this?: kernel: GEOM_ELI: Crypto request failed (ENOMEM). gpt/WD14TB_ZV6C.eli[WRITE(offset=3708987219968, length=1036288)]

I'm seeing them steadily roll in, but as far as i can tell, everything is fine

only unusual thing i'm doing is syncing several TBs of data with syncthing

curious if it stops when that is done

fwiw, it's reported for all 4 HDDs that are part of the same zpool, but not at the same time.

that video sucked. didn't actually talk about clones, snapshots, etc. it was basics of how zfs works but not its features

I generally don't like videos for learning things. I much prefer the written word. Can skim faster to the details I am needing at that moment.

markmcb, Obviously that ENOMEM is sounding like memory stress. But I don't actually know anything. I did not think GELI an expanding amount of memory.

rwp ya the handbook has a zfs page i'm gonna go read that

oh, someone else noticed this too: bugs.freebsd.org/bugzilla/show_bug.cgi?id=275336

Title: 275336 – Processes exiting with SIGSEGV on shutdown on ZFS systems

is there any way to increase swap space on a zfs system if you let the freebsd installer set it up for you?

freebsd's installer only setup a 2GB swap when i have 8GB of ram meaning i can't suspend to disk

Assuming that the installer did the normal thing and used all of the available space then the only easy way is to add another disk.

Or you can backup up everything, reinstall with a different partitioning, then restore from backup.

you can't shrink a zpool can you?

You can make it bigger.

sadly you can't shrink it yeah

additionally you need to move it

you could use bigger disk

that saves you from all the hassle

s/disk/whatever storage this s4-doing system has/

and, unsure if there are good videos on zfs

i haven't needed video for zfs yet

but i've had this problem before, i was able to find text, was given more text upon asking, and noone really got the reason for nor was able to come up with pages with graphics or uni lecture type of video on subject

thanks for the help yesterday, i think its my hw raid controller that is faulty

eh

and that you can't even test

the drives works perfect in other server

same hw, right?

write speeds are ok?

yes its identical server

whoops

have 3 hp bl460c g7 servers

2 with freebsd 1 with debian

wonder if i get correct performance if i skip hw raid from controller and just let it pass through

A month perfectly smooth work on BSD on zfs installed on a dead ssd, on which windows/ntfs struggling to boot, constantly crashing/freezing/bsod...

BSD dont find any errors in /var/log/messages

Is there a logs especially for ZFS to check the filesystem health ?

zfs status clean ( all zeros )

Title: zpool-scrub

nerozero: you may want to not entirely trust that drive anyway, even if zfs does a good job at using it

hello

Where does FreeBSD store sockets for services? Is there a guide like the Filesystem Hierarchy Standard?

orbifx: /var/run seems to be the usual place

thanks vortexx

Also I just found this: man.freebsd.org/cgi/man.cgi?query=h…ath=FreeBSD+8.2-RELEASE&format=html

Title: hier

evilham

thanks !

scan: scrub repaired 0B in 00:00:11 with 0 errors on Mon Apr 1 14:57:49 2024

the purpose of the installing zfs on a dead drive was testing "will it work"? Now - don't even know to say, just wow...

not so dead now, is it? :)

like evilham said, it wouldn't trust it too much though. if you haven't already, you should clone it to a known good drive & put that in your comp.

Any ops around?

mane: in #freebsd-ops

Thanks

ZedHedTed, this is just a ZFS testing machine build, to check how good ZFS is at working with a bad drives, nothing important there.

I knew that ZFS can tolerate and automatically correct some of the drive errors, but working on a dead drive without experiencing glitches at all with 0 errors ...

WOW

Hi guys, quick question: does behyve work on i386, or is it an amd64 only thing? (on the host)

nerozero: oh. very good.

dansimon, this is more dependent on the cpu and platform you got, your hardware should support virtualization

nerozero: it does, I was just wondering if behyve had a hard amd64 build requirement (like some hypervisors do)

I guess the latest builds does, FreeBSD is stopping all 32-bit Hardware support

I read about that somewhere on a bsd announcement site

Really? I did not know that!

Not sure where, it was back in January I guess

freebsd.org/platforms here it is

Title: Platforms | The FreeBSD Project

look at the bsd15

all i386 platforms are not supported except 32-bit ARMv7

hi all

can a (non-vnet) jail be assigned IP addresses which don't exist on the host but might do later?

how do we learn about vnet? only links i found are in the context of jails

but even base networking runs in vnet 0 so it's not a jail only feature i don't think

vnet is specific to jails, there's no other way to use it

the host's "vnet" is just the default IP stack

Hello. Is FreeBSD known to work on Amlogic S905X?

I'm searching the forum now but the answer is no, I won't search

I have this: libre.computer/products/aml-s905x-cc

okay it seems the answer is a big NO

alepzi, Other networking features other than "vnet" such as ifconfig creating bridges and epairs and other things all work everywhere though. That's the place to start. The ifconfig documentation.

Is there any way, these days, to support a Logitech Sphere pan/tilt camera in FreeBSD, or is this thing garbage now?

tmux in freebsd question. outside tmux, i can use my mouse in things like htop. if i run htop in tmux, mouse doesn't work. but i use the same tmux config in other OSes and it works. is there a freebsd+tmux consideration for mouse?

I think it's more the fact that your terminal emulation doesn't support it, but I couldn't tell you details.

Whatever configuration lets it work in Linux will probably work in FreeBSD if you can manage to duplicate it.

i can use mouse in tmux just fine though, it's only the apps in the tmux panes that don't respond

GNU Screen probably doesn't work either.

At least, the mouse.

gh00p, Confirm that you have turned mouse support on in tmux with "set -g mouse on"? You can do that on-the-fly with PREFIX : set -g mouse on Enter

Oops! s/gh00p/markmcb/ for the above! Sorry!

rwp: yep, it's on, can toggle it with prefix-m

I can do all mouse things, it's just not passing the MouseDown1 event to the pane

I disable the mouse in the terminal as it disrupts my use of copy-paste so that set -g mouse on is the best I can suggest.

This should handle it (default): bind-key -T root MouseDown1Pane select-pane -t = \; send-keys -M

but it's not for whatever reason

The mouse will select the pane

but that last "send-keys -M" doesn't seem to happen

same config i use works in Linux and OpenBSD with the mouse event passing into the pane

anyone has a guide on how to find a failing dimm based on these messages: kernel: MCA: Bank 8, Status 0x9c00004001010090 ?

markmcb, If it were me I might look through the Linux & OpenBSD tmux config files and see if they do something specific. Then I would look at patches they might have applied to the code. Then I would try comparing and compiling that version on FreeBSD to see if that works.

rwp: thanks. "tmux list-keys" shows no relevant differences i can see for mouse events. maybe it's how it's built.

it's just odd that it works for tmux, and just doesn't passthrough. seems like if mouse works it'd all work. ... unsolved mystery

Another external input is the terminfo/termcap data for the terminals involved. I don't know but that's just another thought I have of where things might be different.

Anyone seen this? netgate.com/blog/pfsense-software-e…tegic-migration-to-the-linux-kernel

Title: pfSense® Software Embraces Change: A Strategic Migration to the Linux Kernel

it does not seem like an April's Fool joke :(

jmpp, The TrueNAS change is basically the same and definitely already happened well before April 1st.

yeah, for sure. Hardcore TrueNAS CORE user here, so I've been following that with great pain

Also: reddit.com/r/freebsd/comments/1bhvt…e=reddit&utm_medium=web2x&context=3

My interpretation and spin on this is that newcomers are more often comfortable working with the Linux kernel and therefore 1) wish to work with the Linux kernel and 2) more people work with the Linux kernel therefore companies can find Linux kernel expertise more easily and it is easier for people to move jobs.

weh, thats a shame

rwp: yeah, but the drain on other OS' such as FreeBSD is quite a shame

This is motivating me that I need to find more time to become more involved in FreeBSD development.

likewise!

Finished reading through the reddit thread. And it is sad that in the move to OpenZFS that FreeBSD went from being number two behind illimos to being what sounds like an also-ran given that it is said that test cases in OpenZFS don't even run on FreeBSD.

100% agree

that's me on Reddit talking with gonzopancho

he did say they did intend to stand behind FreeBSD at least for years to come, true

but, still, the tone of that Reddit discussion, strongly supporting iX's move, as I understood it, did not instill a lot of confidence on how strong said FreeBSD support will be going into the future

By the nicks I make assumptions-connections. And yes that would be ground truth information about why they are moving.

Does anyone know what he was referring to with "The person who was able got chased out because some reporter at a different big publication decided that clickbait was more important than honest journalism"?

It reinforces that zfs sits as an extremely important part of the system. So important that it has become like a fledgling bird and left the nest flying off on its own with OpenZFS. Which is now a major important part of Linux systems.

vkarlsen, Not I but I assume that referred to some register article. And it had me curious too.

FreeBSD has been a major developer of zfs but now is losing out with OpenZFS due to things like OpenZFS not even running tests on FreeBSD. And FreeBSD kernel struggling to make fixes and improvements that are causing real problems for vendors using zfs in their systems.

it also had me pretty curious, but I didn't manage to follow up on that conversation

the netgate one is probably a joke "linux kernel with freebsd userland" is probably inteded to be a tell

Since the Linux kernel has always had more developers, more churn, it is easier to get changes through that process flow. So they are going to chase the new features which means going to the Linux kernel.

I don't think it is a joke. I think that would be a viable system. Linux kernel + FreeBSD userland.

foxiepaws: don't think it's a joke at all: reddit.com/r/freebsd/comments/1bhvt…e=reddit&utm_medium=web2x&context=3

besides, aren't there already Linux systems + FreeBSD userlands out there? Devuan, or something like that?

i just don't think they'd release something on april fools day like that when yesterday or tomorrow would be equally fine

to avoid misinterpretation with how things are

I think it's just awful timing for the announcement

but, other than that, it rings true to that discussion gonzo had with me on Reddit

It's the reason April Fool's gags work is because for most companies it is just another business as usual day. They make real announcements that we all wish would have waited so that we would not be questioning them.

In that thread it was hard for me to interpret if gonzopancho said something and meant that what they said was good or what they said was bad. For example the reuse/recycle of vnodes. The complaint was that FreeBSD by legacy is a heavy recycle code base but that Linux & Solaris do not recycle. I presume that means they free/allocate instead of recycling. "This is important for the future of zfs." Yes. But in which direction?

I thought it was pretty clear he was heavily critizing FreeBSD's approach

s/heavily/strongly/

I might presume that means that free/allocate is more dynamic in the use of memory and therefore perhaps can handle short duration memory stress spikes better? But that does not mean that recycling excludes using more dynamic memory too. I just don't know enough about the code base.

I don't know any about that code base, hence my initial reluctance to reply in any depth

I could see that recycling would be much more efficient. Better cache locality. And free/alloc often suffers from what I will say is memory fragmentation due to lack of a better name if memory chunks don't stack up like tetris as nicely and holes of unusual sizes get generated. Not saying there isn't techniques to make that work but it's not absolutely trivial. Look at the effort put into malloc() over the years. It's not a trivial

problem.

also raw speed, 'cause when you allocate a new memory chunk, you have to find said chunk to begin with

whereas if you recycle an already allocated chunk, you spare yourself that potentially expensive step

Sometimes the grass appears greener on the other side of the fence. Sometimes wanting something is not as pleasing of a thing as having. This might be a case of that happening now with these vendors who are jumping from here to there. In which case some might return. Some might continue to offer both.

monocultures are never a good thing

and Linux is very quickly becoming one

jmpp: Ed says: no EN/SA for libarchive #2101 but perhaps for #2107 mastodon.social/@emaste/112197323460302957

Title: Ed Maste: "@wednesday⊙nhi IMO the fix in #2101 is too…" - MastodonEd Maste: "@wednesday⊙nhi IMO the fix in #2101 is too…" - Mastodon

Speaking of someone who probably has not had much sleep the past few days, emaste has almost certainly been on task saturation. Can't thank him enough for his work and effort on this.

"IMO the fix in #2101 is too minor to warrant a EN or SA." nice!

Just it by itself seems low risk probability too. That printing of unescaped file names feels like something for a yet different serial stream of connections that the malicious actor was putting in to have available for something else. It would need to be chained along with other things.

but definitely a very dangerous step to have let slip through

thankfully it was caught!

pretty sure that one wasn't malicious, there are a bunch of other instances of not escaping that are pre-existing

maybe not malicious in itself, as in, not leading to an exploit directly

There was a patch put in where there was one review comment before it was merged that said only lgtm (looks good to me) and then it was merged. I looked at that patch and there is no way I would have connected it. Especially since it was *six months prior* to when what the patch affected was injected into other parts of the code base. At the time it was not malicious. But it became malicious later. Truly genius.

but if created by the same xz actor, then it wouldn't surprise me it was in preparation of

exactly

Playing the long-game there with the patience of the sphynx. It's impressive in its cold and calculating malevolence.

there was commit, IIRC on libarchive precisely, and IIRC by this same xz "person", where someone made a comment to the effect of "if not malicious, certainly on track to be"

precisely the point

i've been wondering about this sort of attack ever since the "FBI put a backdoor in OpenBSD IPsec" thing - which i guess turned out to be false in the end, but the idea is clearly viable. it's been done before at least once by researchers: arstechnica.com/gadgets/2021/04/lin…ty-of-minnesota-researchers-apology

Title: Linux kernel team rejects University of Minnesota researchers’ apology | Ars Technica

i'm curious if there have been any other notable malicious examples

You mean like Ken Thompson? Reflections on Trusting Trust by Ken Thompson genius.cat-v.org/ken-thompson/texts/trusting-trust

Title: ACM Classic: Reflections on Trusting Trust by Ken Thompson

well, there's Ken Thopmson's paper, of course

I think that is the one. I just searched for it a moment ago to cite it.

i was thinking more like real world examples

Real world examples are sometimes not found, not caught, still available to be active, and so we just don't know.

lw: Thompson's point was, precisely, you could not tell if his was actually a real world example... or not ;)

kevans: libarchive/libarchive 02cfa8a#commitcomment-140401844

Title: Reusing code from zip size known and adjusting comments · libarchive/libarchive@02cfa8a · GitHub

i am familiar with the paper, i intended to ask if there are any notable examples *that we know about* (since if we didn't, they wouldn't be notable...)

yes, i was eyeballing that a couple days ago

different issue, perhaps even different code base, but I'm referring to the potential long game pursuit here

lw: certainly lots of finger pointing and even accusations, even about RSA encryption, but right now a demonstrated real world example, other than xz of course, doens't come to mind

Hadn't Jia Tan's GH profile been disabled?! github.com/JiaT75

Title: JiaT75 (Jia Tan) · GitHub

i think github received some flak for disability the repository/users involved since it made analysis harder, maybe they undid that

s/disability/disabling

yeah, they did, but xz's is still disabled, though

anyone knows how to interpret MCA errors ?

trying to identify a failing dimm

last1: if possible check your ipmi event log tool

e.g. ipmitool sel list

last1: dmidecode might have per-socket error counts

there's no errors in ipmi

last1, I admit to brute force last time I had this problem. I removed half the ram at a time and ran memtest86(+) and did a binary search to identify the bad ram. Which turned out to be the dimm socket on the motherboard! Drat!

lol

that's pretty hard to do on a production system :|

I was hoping to locating the correct dimm and getting it rma'd

mason: for each dimm ?

looking

I think for each socket.

but I could be confused here

I'm checking too, I don't see anything like that for the dimms

Title: Handle 0x0020, DMI type 17, 84 bytesMemory Device Array Handle: 0x001 - Pastebin.com

I see the dimm says: Bank Locator: P0_Node0_Channel0_Dimm0

can that be mapped to MCA banks ?

last1: Is that just one stick? And is it ECC?

that's one of the sticks yes, I have 4 x 32Gb, ECC

last1: Have you looked at this yet? freshports.org/sysutils/mcelog

Title: FreshPorts -- sysutils/mcelog: Collects and decodes Machine Check Exception data

I could have sworn I got memory error data out of dmidecode in the past, but I'm not seeing it just now.

yeah, I used mcelog too, didn't give me more info sadly

last1, I looked in my notes and see a util-linux package reference to lsmem and chmem there. "lsmem" and "lsmem -a" and a note that ,,With "chmem" you can temparily disable memory regions at runtime. So after you found out which RAM bar it is, you may test your suspicion by removing the specific RAM region with "chmem".''

Ah, "error information handle" in dmidecode might be what I'm thinking of.

I know, util-linux, but it's related information. Maybe some helpful information there regardless.

so what would the point of this warning be in its current format ?

just to get you to start reseating ram and checking them one by one ?

I wonder how this guy mapped his error to dmi decode bank 8

That second comment about swapping dimms in slots is one I have used before to identify errors with locations.

ah lol

it says in his dmidecode: Bank Locator: BANK8

hang on a second, maybe it's on the motherboard manual

P1_Node1_Channel0_Dimm0

or not, seems my issue is somewhat common and nobody knows how to figure it out :( truenas.com/community/threads/findi…ion-on-supermicro-x10drl-i-mb.78078

Title: Finding bad memory chip location on Supermicro X10drl-i MB | TrueNAS Community

last1: That's the thing - I remember a while back seeing recovered errors, and I thought where I saw them was dmidecode. But I have none showing now so I can't confirm itk.

s/itk/it/

twitter.com/hnasr/status/1774658928913387992 kind of on topic of what we were recently discussing

jmpp: that's bad indeed, but playing devil's advocate, is that much worse than willingly giving up your data to say CloudFlare ?

they are basically mitm, but paid

heh, very very true!

still, there's the issue of consent

even if in the case of CloudFlare it's objectionable, it's still consensual

whereas in the case of Facebook... eehhh...

I honestly don't think many people understand that CF terminates their SSL connection

especially since their free tier is...free

attracts a whole bunch of clueless people

is reddit down for you guys ?

I get 503 Service Temporarily Unavailable

not in France

failed for me a couple of times with 503, indeed

failing indeed right now

I understand the Linux kernel is very dynamic and well maintained and blah blah

but the distributions, really, what a shitshow

systemd is a nightmare

10 different ways of doing things across 10 different distributions

last1: yikes, yes, they kicked and forced me out of the Linux ecosystem, I fled in horror to the harmony and consistency of the FreeBSD userland

last1: and please do not mention CentOS "security" framework monstrosity, I'll suffer a PTSD episode!

selinux?

you're going to cost me a call to a therapist!

:P

Title: 'CVS: cvs.openbsd.org: src' - MARC

Yikes, I always thankfully forget that OpenBSD is still using CVS!

"xz is a lifted Ford F-250 that "just ran a stop and killed a cyclist"." was that said in hindsight? Or was it visionary...?

:D

wait, you do know about the xz incident, right ?

of course, been talking about it here all weekend, hence my comment

:)

wait, I'm just noticing the date on that commit

is it an April Fool's joke, or something?

the bigest indicator is probably "in good company... kernel.org, FreeBSD, and Debian"

there's no way that they'd consider us good company :-)

heh!

I am surprised an April Fool's joke would extend to such a large and potentially misleading commit

Dschihad(-75) in Ramadan!

That is very unreligious.

marc.info/?l=openbsd-cvs&m=171200380712333&w=2 first respose

Title: 'Re: CVS: cvs.openbsd.org: src' - MARC

tbf it's almost certainly just a fake e-mail

yeah, that's what I came to think, but man, was that crafted!

we have soem good ones on occasion, too

lists.freebsd.org/archives/dev-commits-src-all/2023-April/024787.html was pretty good, haven't seen one fly by this year

Title: git: f001edfb5d41 - internal/admin - Add Charles Root

on the dmidecode, after reading countless pages, there's no reliable way to do it except through manual trial and error

so instead of doing that I'm rma'ing all 4 sticks

and let Samsung figure it out

end of story, too much time wasted on that

last1: Good call honestly.

out of so many, one phrase Bryan Cantrill once said about sys developers separating from sys admins when the time came to understand why a system had crashed. The former don't want you to touch the system, at all sometimes, as that may alter the evidence and impact the investigation; the latter, instead, just wants to get the system back up and running ASAP!

and I loved that comparison 'cause I do a lot of sys admin, but I'm also a dev (though not a low-level OS dev, perhaps unfortunately)

and that mix definitely helped me see the light and learn to classify problems, out of the so many ways they can be classified, into two categories: 1) those I *care* to understand and try to solve, and 2) those I *don't*!

someone pays you to care ?

to bring the platform back up ASAP, for the most part, but then my OCD to understand why something is not working as expected tends to interfere

meh, I moved past that a long time ago

family is more important than the company

very very true

but I'm also a physicist by profession, so it kind of comes built-in with the brain wiring

so quantum physics and those dice must really mess with your brain then

let's just say I had high hopes for Einstein's unrelenting position that, at the end of the day, the world *had* to be deterministic :P

I still can't imagine the fabric/geometry of space time as posited by GR can morphed into gravitation as a field

it's profoundly at odds that mass generates the shape in the fabric, vs the graviton and its field

hopefully they can advance a bit during my lifetime

last1: I don't quite like the "fabric" moniker, I think the morphing of spacetime by a mass/energy distribution is much easier understand if you envision the universe as some kind of a foam in which we're embedded

so, as it's hopefully easy to see, if you step into a massive bubble bath, your mass distribution is going to disrupt that foam around it, in a way intrinsically linked to your specific mass distribution

and that's easy to understand

but different than me emitting gravitons as I enter the bubble bath

on the other hand, gravity waves have been detected, so who the hell knows

well, it turns out every mass-energy distribution does the exact same to the universe itself, and the rest, what we perceive as gravity, is just the Hamilton's principle of least action, i.e. things will evolve in real life in the way that requires the least amount of energy

as in, you naturally buying the product that cost the least while still meeting your specific requirements

well, that's essentially what a geodesic is, and moving along those is what our minds perceive as "falling in a gravitational field"

yep, that's GR

when you talk about gravitons, you're getting back into that pesky QM! :P

how can i verify that i DONT have any swap space? sudo top doesn't show it for my host machine but all of my vms do. is that proof?

alepzi, "swapinfo"

tried that, shows nothing

Then you don't have any swap configured.

so i guess that's no swap? but my bsdinstall installerconfig has ZFSBOOT_SWAP_SIZE="1G" like all my vms have so wtf? it has ZFSBOOT_SWAP_ENCRYPTION=yes but the vms don't, is that breaking it?

alepzi, Example from here showing swap configured: termbin.com/cvu7

ya my vms show that info but host doesn't

alepzi, Just as extra information perhaps run "geom -t | less -i +/swap" and look for the swap entries partitioned?

so ZFSBOOT_SWAP_ENCRYPTION=yes is making swap not be created?

hmm

that command shows a swap label and dev on each of the mirrored drives

What do you see here: grep swap /etc/fstab

Another example from here: termbin.com/fhjl

2 lines

i have the same, /dev/nvd0p2.eli and /dev/nvd1p3.eli

If swap is configured in /etc/fstab but not present on your system then I would try "swapon" and see what errors are emitted when it is attempted.

same values for each column

Sorry, "swapon -a"

said swapon: /dev/nvd0p3.eli: Invalid parameters, same for /dev/nvd1p3.eli

Hmm... My 13.3R recently upgraded system has no swap either. Did you recently upgrade to 13.3R?

ya i'm on 13.3

but my vms are too

My recently upgraded /etc/fstab is zero sized. Hmm... Looking through the old snapshots to see when that changed...

termbin.com/c4yx is my host's ZFS config in installerconfig, just to verify those settings are right?

To show my ignorance I don't know how to drive the installer with those variables. It's not something I have done.

ok np

What does "gmirror status" show for you? On my system the mirror is active. But /etc/fstab is zero sized so it is not mounted. I need to fix that.

status not available, try load first

And the crazy thing is on my system it looks like it has always been that way from the earliest snapshot I look at. So it must have been an oversight on my part on setting up that system. Time to look at another one.

Oh good. The rest of my collective is okay. Just something I missed on that one. That's a good find.