right, so you're probably still descending into other ports that depend on them

ps -d may be somewhat enlightening here

kevans: sorry for the delay, fell asleep in my chair... ps -d shows a LOT of output. It's a tree so I can only see the command for the first 15 or so. There are 25077 lines in the file I piped the output to...

The top-level make command seems to be py-recommonmark

de

demido: what was the cause of your swap

no clue

tried a 3rd time, this time with a separate partition for the vm images. same prob

so now i'm trying with 2 raid arrays. 2nd array mounted and holding the vm images

we'll see how that goes

if it doesn't work then ufs is pretty much garbageware and i don't think i can use freebsd because these servers have perc 730p mini raid cards and i don't think those have a good hba mode meaning zfs is out and i gotta seriously consider using debian instead sadly

Update on my infinite make install loop thing... I pressed CTRL+C to end the loop because I'm pretty much bored with it now. It stopped, and then started back again. I held CTRL+C down for like 30 seconds and it's still going.

it shows the PS1 prompt and everything

I can even run a "echo howdy" command, it outputs just fine, but then more make-related stuff keeps coming through

And the last bit of intel before I restart this thing... It used 100% of 64GB of ram, and running "top" took like 3 full minutes to open on a TTY.

i don't see how using raid is going to fix a memory problem

well ya it prolly wont rtprio but i'm trying everything before i'm forced off of freebsd and onto linux

you think -current would do better or nah?

why is there always 724 users here

Aubriella: people using bouncers probably

Aventra, yeah, but wouldn't you expect some variance there?

lol

thats a global population

supposedly

whats the variance, it's at least seasonal

with a drawdown of a few people an hour

ya using a 2nd raid array didn't help. back to 100% swap used, swp_pager_getswapspace failed, out of swap space

don't see any way i can stick with freebsd this sucks

Have you read through this? man.freebsd.org/cgi/man.cgi?tuning

first sentence recommends 192GB swap, since that's how much ram i have. that seems reasonable to you all?

so i put the raid card into hba mode and did a fresh 14.2 install with auto zfs and mostly defaults, including using local_unbound. but when i boot into system it says error resolving pool 2.freebsd.pool.ntp.org name does not resolve

can't resolve any domains

i selected dhcp settings when installing and i can ping out. i can even ssh in, so i know it has network

i thought the recommendation is double RAM up to 4 GB (so 8 GB swap) but over that one might never need it

'double (of) up to 4 GB RAM'

what's vdev i/o failure mean? zpool=zroot path= offset=137472937984 size=4096 error=5 getting tons of those in /var/log/messages on freebsd

i checked the smartctl output of the drives but i don't know how to interpret it. there are some non-medium errors for each drive fwiw

demido: vdev i/o failure is bad

error 5 is EIO (input/output error)

your hardware (controller, or disks) is failure)

s/failure/faulty/

drop output of `zroot status -v` here

do you mean zpool?

#facepalm

zroot no command found

`zpool status -v`

demido: aye I just got up, ENOCOFFEE

demido: IIRC the original plan was:

dch termbin.com/t5sz

- switch to zfs, because we can then change the memory allocation for caching of the VMS

ya

i can't get a break heh

mfisyspd7p3 is the one with the issue

its a single checksum, a single write failure

how could you tell that's the 1 with the problem?

so do the smartctl checks on this, stash the results somewhere for safekeeping

then we will clear the error flag, run a `zpool scrub zroot` and see what comes up

sorry my copypasting is bad today

mfisyspd5p3 ONLINE 0 435 1

far column

ok ya that's what i thought

so do smartctl check on just that drive and save it?

this entire system is cursed

the WRITE column is *also* an error counter

so every single disk is bad?

mfisyspd1p3 has had 1.03k failures writes

demido: possibly

fuck

its more plausible that this is a RAID card, and the volumes are hardware raid, and zfs is cleaning up issues there.

demido: any more detail on the hardware?

perc 730p in hba mode

do you want me to save the smartctl output for every drivethen do scrub thing?

you should use the mrsas driver in preference here I think

instead of using hba mode?

or use mrsas driver WITH card in hba mode?

no, I need to get some coffee/food first, but mfi driver is old, freebsd 6ish, and mrsas man.freebsd.org/mrsas is the newer one freebsd 10 IIRC.

no to which?

my first guess on seeing this is that that h/w card is in some sort of RAIDy mode still

or the card is defective, thats a lot of write errors for a system that you just installed

so we should be aiming for mrsas drivers, using "JBOD" mode.

I'm not sure wht HBA mode is, probably the same thing.

NGL this is a little fiddly

I am not sure I have access anymore to a system where I did this, i'll have a look after breakfast

is it a good sign that i run zpool status -v again and the write and chk numbers aren't higher?

demido: if it can't survive the installer, its going to shit itself under load

well i've installed and configured the 2 vms on it fwiw

overall, you need to add stuff to /boot/loader.conf to tell mfi driver not to connect, and to load the mrsas one too

first lets clear the pool state, `zpool clear zroot`

then `zpool scrub zroot`

you can watch it via `zpool status -v 5`

ok

get that coffee and i'll be here

does the current FreeNAS still use freebsd?

scrub repaired 10M in :38 with 0 errors

in the output it's showing 0's for all write columns, but read and chsum columns have lots of numbers

demido: ok so this is definitely not viable setup then

ok i'll config it to use the mrsas driver then reboot it

that wont IMO fix these errors

damn

I only have notes from a DELL-BOSS system, but we remove all the "virtual disks" in the bios

my best guess so far is (if this is a new system) that the hardware is still in synthetic raid mode somehow

used

and if its not a new system then either controller, or cables, are defective, or not well mounted

sometimes its enough to open the box, wiggle the cards and cables and reseat them

anyway sadly I'm afk for most of the day now, I need to pack & wrangle kids

good luck demido

damn, k ty

once your h/w issues are sorted

`zfs set primarycache=metadata zroot/my/vm` was what we wanted to do, for memory tuning

i dont see how ill fix them

maybe the newer driver is enough

is it true I must delete either KDE6 or XFCE4 (conflicts) to upgrade with pkg?

pretty sure they don't conflict directly, so conflicting dependencies?

yay virt-manager and bhyve

now i just need something better than xrdp

dch im setting the box up from scratch again, taking exact notes, and using mrsas driver. so far so good, no errors anywhere (OMFG?)

dch got the system fully rebuilt and now the vms are downloading data. will take a few hours to finish that part but it's looking great. all zeros in zpool status -v 15, no errors in /var/log/messages

i set `zfs set primarycache=metadata zroot/my/vm` so hopefully the mem issue is worked out now too

in my zfs config, should i enable swap mirroring? i feel like i should since losing the drive with swap on it would take the system down, but is there a downside? like why is it disabled by default?

Hello. Here's the situation: I have a hp proliant g8 server and I have freebsd installed in bay0. but i want to copy the OS into an SD card so i can have an extra drive for the RAID. So i've found the best thing to do is to use an SD Card. what is the best way to copy the operating system from the hard drive to the sd card, considerating the sd card is smaller (1.2b of hard drive vs 128gb of sd

card)

currently the OS is only 3GB. So it's not a big issue

Anyone here at FOSDEM?

unfortunately not for me this year :(

polarian: have fun though :D

Thanks :)

I'm no expert but I

oops

I'm no expert but I can't imagine this is correct.. Everything I try to install through ports keeps wanting to pull in way more dependencies than it seems like it needs. Like why does Bash and porttree and nano need all this python linear algebra stuff?

well, porttree is written in python using "all this python linear algebra stuff"

hello all, i'm asking a question in #freebsd because i am dealing with a descendent that uses y'alls drivers. what is the state of usb ethernet dongles in freebsd? e.g. consumer grade like belkin. in openbsd it's a hot mess, very finicky and unreliable. i'm curious whether freebsd has this ironed out pretty well. (mine appear as ue0 and ue1 but that is all i know)

(i'm at phase where i need to use what i have in the junk drawer or purchase new ones)

to be honest i only have 1 that appear and one that needs modeswitch

the appearing one is untested :/

:p

wifi has fancy improvements tho if you want that

it's really a try-and-see thing but i've been bitten by getting a deluxe network enviroment all set up only to lock up at some stochastically determined time

eh

ketas: i have a little nuc that sits next to router. i'm trying to turn it into a mini-multinic system with pf rdr-to/reply-to to create symmetrical routing

messy messy messy hmm

ketas: thanks, i'm afraid to read that because i think the answer is discouraging

any chance you can do what i did?

switch

but did you try

switch like network appliance or switch like change dongle

ketas: i have only confirmed that i can get ue0 and ue1 to get dhcp and receive traffic

but the pathologies i have seen are in longer time, like in a day or two it stops working

switch like managed ethernet switch, they appear as second hand items, 1g should be pretty available etc

i'm really at the point of maximal economy and sport to be had from keeping out of landfill

if i really wanted to do this i would but multiple e1000's for example and a new computer

but you had some ethernets around

so i would put $ on switch toward proper solution. but i don't want to spend the money since this current equipment would go in cabinet

what happened?

ketas: wrong form factor, i have a nuc. the ethernets were for a homebuild from 2017 that died

wrong form factor for usb ethernet?

sorry, i have two usb ethernet dongles (ue0) and several network cards (Intel (TM) e1000) but no computer for them

you could connect them to each other with cat5 and flood something through

for a day

i had idea to use them but somehow didn't

indeed. hoping a little check in with #freebsd might help me decide to put weekend into this hack or whether i need to head to microcenter

i bet someone uses them

it *did not* work on openbsd. it would not be an option. i am hoping for different outcome by different driver implementation

all i know is that many of those drivers are been there in fbsd for long time

unsure what it means

either work or not

i hear you

it could be really good and stable or really bad and shitty

forum topic confirmed working of some or so

unsure

ketas: i think i've been here before years ago with pfsense, lan & wan on ue and it was poor experience

so, knowing that drivers areold helps me believe that situation is not changed and i should get new chassis for e1000's

thanks for feedback. i'll keep my eyes peeled for more feedback

ecbrown: when i pointed out weird usb wifi delay/latency/jitter/loss issues to adrian, he was like maybe indeed usb stack itself isn't really that realtime perf that nics need so if your usb eth issues with every card was like this, who knows

also i don't think i'm exactly that good in this at all

but i guess they are rarely used

that's bummer

like e1000 is everywhere, god forbid if it's not working well

well, i have bigger problems too, i have 10 gig available to me.

Hey do you guys think it's possible to run FreeBSD on a 5k iMac with the AMD GPU available?

but i really need to plan my investment. i don't mind spending money but i want it to be what i need

Aventra: wiki.freebsd.org/AppleSilicon

did you see this page

no - 'tis an intel imac sorry

then i'd say your chances are good. who knows about wifi, plan on a dongle

my main concerns with it would be fusion drive, wireless(I could use ethernet I suppose), and graphics through the gpu

the best thing to do is try and see with a live stick

Aventra: you will have to break up that fusion drive

or perhaps do a deal with the devil in volume manager

yeah that's fine - I think you can install root partiton to the SSD and use the rest as storage

forget about the spinnning rust

dd target for backing up ssd portion

plug in ssd additional storage and that thing will rock

when i had one i was sooooo glad i got the 128GB

ecbrown: 10g broadband0

?

i used it so long that the hdd (even unused) developed an audible quack

i had to ditch the computer because i screwed up taking off screen to get at power connector

ketas: yes

might as well use mellanox nics then

they have their own fbsd driver writers

ketas: that's what i'm talking about

(dog)

yes, certified high speed ethernet cards

unsure what netflix uses in their fbsd boxes

at 10g one needs like actual hw too

not those soda straws

yes, this is why i'm migrating from openbsd as well. sometimes it has a hard time keeping up with that speed

i love that os but i have to with what works

s/go with/

i switched my setup to zfs and it seems to be better, but i have 57G Free yet swap is 46% used? 8G laundry too

memory use it or lose it

if it evicts who cares

i use hammer2 so the memory in use is lower but its only catharsis

ecbrown: I can never find anything about that. Does Hammer2 do self-healing if it has multiple backing disks yet?

no

it is merely a filesystem and not a volume manager

i'm not an expert but i feel like its btrfs or something in capabilties

I thought that was one of the original design goals for Hammer, but I haven't tracked it closely enough.

Was also a design goal for Stratis, but that didn't happen.

iiuc a motivating goal was for "federated kernels and storage"

i am so old that we had something called MOSIX that did this

but now there is exptl support for mounting remote hammer2 volumes, that is not e.g. nfs

yes, that is my understanding

btw i am not here to litigate for dfly in any way. i love zfs and freebsd

I was just personally curious.

imma now say something controversial, i love ufs and ffs. sometimes i just dont need the extras

heh

ecbrown maybe you didn't read what i said, but swap is being used when i have 57G FREE MEMORY

so what

thats what i wrote to what you wrote

why second guess virtual memory system?

demido: It's not being actively used, is it?

demido: I'd suspect this is from some earlier event.

i would like to ask another topic: i grew up on mvs and so have familiarity with enterprise solutions that are way past what you can get at best buy. how does freebsd + zfs going full tilt compare with e.g. what a IBM Power storage offering provides today?

is it basically comparing infinity to infinity?

(1 petabyte for argument sake)

gcp provides moar :D

ecbrown: I could be wildly wrong, but my understanding is that enterprise storage outside of ZFS or things like OneFS or WAFL you basically have Ceph or Gluster.

there is no ibm these days

wow, did not know any of this landscape. glad i asked

ecbrown: Also, contrary to what you just heard, there's absolutely still IBM and Power in big enterprise.

hah

thank you for discussion, #freebsd. i'm new to your family group and glad i took a spin by. i hope to submit some patches to freebsd-ports that will benefit me downstream. cheers and great afternnoon

mason not being actively used. it's free, not active or inactive or laundry. 57G free. swap is 43% inuse and really shouldn't be with so much free mem

mason but now i'm using zfs, so maybe it'll make it easier for us to figure out why that is?

demido: It won't probably go away without some reason to. Nothing much would be fained.

gained*

seems lame to use swap when there's 57 GIGABYTES of free mem tho

demido: Force it out if you want. swapoff, swapon

But it won't gain anything tangible.

i'll leave it alone i trust your knowledge. can you tell me how to detect when swap usage is BAD though so i can watch for that?

demido: I'd worry about it if you get to where main memory is full. Until then, any use of swap is generally incidental.

ok ty

Sure. And again, I don't know why it decided to dig into swap a bit, or when, and that's certainly a curious thing.

But what's bad about swap usage is that it loads an already busy system with unwanted IO.

look at the swap rate not the fill level

how can i do that?

find a utility that reports page-in-per-second. On Linux it'd be vmstat

q

vmstat 5 is running now

pi/po started at 7/5, then 0's each update since

just got 2 pi, then 1 pi

in my terms, that's as close to idle as to not care about

just got 3 pi

just got 9 pi

still np?

yup

just got 4 pi

ok

what kinda numbers would be worth taking note of?

thousands, depending on your hardware

ah ok great

and in the auto zfs screen of bsdinstall, i'll enable geli encryption, swap encryption, mirror swap, and raid10, with default 2g swap size. any other settings important on that screen?

Does QT make use of the system CA? It doesn't appear to.