V_PauAmma_V: try it and see. i don't see why not

elliot@phil:~$ sysrc -f /tmp/blah foo.bar="baz"

sysrc: foo.bar: name contains characters not allowed in shell

bah

now that I recall, I already ran into that one, and obviously forgot about it

forgeg about ig

bah im trying to select an iso image and it wont let me with cbsd bconstruct-tui

any idea why

i can run the command and the interface opens, but when i hit enter in the "vm_iso_path" it attempts something and goes back without letting me change the iso image

i can't do it myself, but it's easy! bugs.freebsd.org/bugzilla/show_bug.cgi?id=270530#c2

Title: 270530 – sysutils/iocage: missing jail parameters

freshbsd.org/freebsd/src/commit/e315351fc7af11 woo!

Title: FreeBSD / src / e315351 / Add the mfi(4) ioctl support to mrsas(4) - FreshBSD

Freaky: looking at freshbsd.org, "committer" selection says "100+" while actually listing 1000+ (i understand that as commits count?)

Hmm, apparently there are restrictions as to the name of a netif when attaching to a bhyve vm?

bhyve kept giving me, "device emulation initialization error: No such file or directory" when trying to use tap interface "vm_debian-11"

Worked when I changed it to tap0 though. Wonder if it's the hyphen or something.

_xor: can you please submit that as bug

zlei is currently working on making renaming interfaces more consistent and less… fraught

Is there an on-going issue related to this or is it going to be a new one on bugzilla?

try truss'ing it first and see what exactly happens

_xor: there's a bunch

might be good to collect them in a meta issue actually

Busy at the moment, but can do it later today or tomorrow. I have two issues I can raise, this one with tap as well as one regarding netgraph node name limitation (which was undocumented last I checked, so it should either be documented or the limit increased). I noticed that bhyve can use netgraph nodes, and since I already use netgraph, I might

use that instead of tap interfaces.

yuripv: "100+" just means the facet is showing only the top 100 with everyone else grouped as "Other"

and yes, it's commits matching that facet

if you click on the heading the facet will expand, so your "100+" becomes "980" for all FreeBSD committers

that's a lot of committers

I reckon we may have a few more authors than committees

committers, even

I wish SVN and CVS had had that concept

contributors? yeah, there's a lot of them

docs.freebsd.org/en/articles/contributors/#contrib-additional here's a list that's been built up over the years

Title: Contributors to FreeBSD | FreeBSD Documentation Portal

good morning everyone

freebsd is so awesome

finally have bhyve working the way i want it

:)

I've got FreeBSD 13.1p7 running on an HP mini-pc type system. FBSD works fine, but even with no activity, the fan is constantly running. Is there a way to control that in FBSD ?

neither htop nor btop show any CPU related activities nor any I/O activities

librehardwaremonitor could presumably be integrated into a kernel module, which could live in ports, that could present the devices for a userspace daemon like bsdfan by claudiozz - but i don't know of anyone who's working on that

trying to grasp something here. testing write speeds to disk. so doing a simple cat /dev/random to file.. how is it possible for the file to grow in size after i kill the process? I assume that it is data getting written to disk from cache on zfs?

are you trying to benchmark zfs or the disk?

your comment implies both, but that's not how benchmarking works;`diskinfo -cit <device>` will benchmark individual harddisks, but you'll need something like benchmarks/fio in order to generate synthetic filesystem loads

synthetic filesystem workloads also aren't very useful for realworld performance, so it's doubtful you'll get much out of it

Hi, I'm trying to use mpd5 (as a client to a pptp server) from a Jail and I'm getthing "MppcTestCap: can't create socket node: Operation not permitted". You can see my config here: forums.freebsd.org/threads/can-i-use-mpd5-in-jail.13925

Title: Can i use mpd5 in jail? | The FreeBSD Forums

martinrame: you're probably missing some devfs.rules(5) combined with a ruleset in rc.conf(5)?

My best suggestion is to use truss (or dtruss from sysutils/dtrace-toolkit) to find out where it's failing and work your way from there.

debdrup: thanks. I'll try to see what happens.

debdrup: I the issue posted here: forums.freebsd.org/threads/can-i-use-mpd5-in-jail.13925 I attached the output of truss, there are some "Operation not permitted", but I cannot see how to fix those.

Title: Can i use mpd5 in jail? | The FreeBSD Forums

Looks to me like you haven't loaded all the relevant modules.

mpd5 normally does this, but since you're in a jail, it can't.

debdrup: I thought it should be possible in a VNET jail

There's no kernel in a jail.

Mm, I don't want to connet to that vpn from the host, any alternative?

s/connet/connect/

martinrame: find out what kernel modules are needed, load them, and present them to the jail via devfs.rules(5) like I described before.

debdrup: thanks!

among others, when I run truss mpd5 I get: kldload("ng_socket.ko") Operation Not Permitted

How can I add that module to my devfs.rules?

I'm not sure jails can load modules, you might have to preload it before the jail starts

Deja vu.

I didn't scroll enough

I think you need to speak in smol sentences

ˢᵐᵃˡˡᶜᵃᵖˢ ⁱᵗ ⁱˢ

martinrame: first: man.freebsd.org/rc.conf(5) kld_list

Title: rc.conf(5)

debdrup: yea, what I really am trying to do, is to test the writespeed to my zraid1 setup.

so, yea, im trying to bench zfs with my current disk setup.

drobban: what are you trying to accomplish with those benchmarks?

except from figuring out writespeeds?

What do you hope to learn from knowing write speeds of a synthetic workload?

im trying to learn what the writespeed is.

dude

Yes, and I'm asking what you think that'll help you with.

they'll probably be roughly the same as when you calculate it

meena: well as for now, the calculated writespeed and the actual write speed differs.

by how much?

80%

=)

80% difference or 80% of the calculated speed?

so Im probably calculating it wrong or missunderstood how zfs is working with the drives

Also, what sort of technology is backing ZFS?

zraid1, freebsd, disks with 500MB/s write speed

64gb ram, ryzen 9.

Is that according to the specification sheet, or from something you measured on the disks?

from spec.

Right.

SATAIII is 6Gbps which works out to ~550MBps, but that doesn't account for 8/10bit encoding, so the real bandwidth isn't as high as all that.

I would recommend using `diskinfo -cit <device>` on the actual raw devices before you create the pool, to get real-world speeds rather than specifications - since they're basically never true.

yea, is there a "correct" way to meassure the pool?

benchmarks/fio like I also mentioned before.

debdrup: thanks, will take a look

The better way is to do custom tooling that matches your realworld workload.

debdrup: for now im just interested to see where the bottleneck is.

with a pool with three disks in zraid1, I would assume the load is split onto 2disks and 1 disk for "parity".

but that isnt what Im seeing. But will test fio

That's not how raidz works, no.

The stripes are written to all disks in a raidz vdev, and parity is distributed.

yea. but how does the calculation differ do you mean?

Err.

The stripes are written to each disk in sequence in a raidz vdev, and parity is distributed across all of them.

It's done via an XOR, the details are here: people.freebsd.org/~gibbs/zfs_doxygenation/html/da/dc9/RaidZ.html

in sequence?

Title: FreeBSD ZFS:

debdrup: yea, I know how "xor" works

I'm not sure what calculations you're talking about.

yea, the maximal bandwidth for one disk is one thing. But writing the total load onto multiple disk in parallel gives an entirely different number, and sure. one third of the data is the sum of the xor operation that can be used to restore "lost" data

I think you need to forget what's on the specs, and instead measure what you get out of it.

debdrup: yea, that is what Im trying to figure out how to x)

By measure what you get I mean use diskinfo and fio.

drobban: do you have any data on that pool that's of value?

meena: yes =)

does zfs send | receive have the potential to show maximum transfer capabilities of the pool ?

I'm not sure what you're asking.

last1, Yes. send | receive should be the fastest way to transfer data. If the transfer is over the network then of course network bandwidth limits apply.

I'm over a 10Gbps network and I'm transferring from a 6Gbps enclosure ( 72 x hdd ) to another enclosure

hitting about 2.6Gbps using multiple rsyncs

although the limiting factor there is the 6gbps backplane

and possibly the rsync method. I'll see if zfs send/receive is faster

The main difference is that rsync needs to deal with individual files. It will be open, read, close, open, write, close, file by file.

Whereas zfs deals with it at the storage block level and with the aggregate of the raw data.

also rsync does a lot of other stuff depending on options.

That does not mean that zfs send|recv will always be faster than rsync. I can see some conditions where multiple parallel rsync can be faster.

My point is that they are operating at much different layers of the system and will have different benefits.

With ~40 TB of data on SATA 3 disks in RAIDZ3 (forgot the number of disks or vdevs) over 1 Gb/s network took ~16 days with zfs-send | zfs-recv; rsync took ~17.5 days

So overall that was a wash

hmm

40TB (assuming binary units) is about 35*10^14 bits

The transfer data -- size / hour -- followed a quadratic regression after 3 days

Sorry cubic

oops, 35*10^13 bits

assuming 700Mbps to make the calculation easy, that's about 5*10^5 seconds

which is just under 6 days, so the network wasn't the limiting factor

"Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway." --Andrew S. Tannenbaum

RhodiumToad, you made me to check my fuzzy memory; will update if I would find the numbers ...

Oh I guess that is a paraphrase of a previous saying that was already in common use at the time. I will amend my attribution of it in the future.

now imagine a cargo plane full of microSD cards :-)

I started this transfer about 3 days ago, but it's not always full speed @ 2.6gbps due to power constraints

the 72xhdd draw a lot of current during business hours so I keep it lower 9-5

but so far I did 22Tb

another 100Tb to go

I'll stop being OT after this one: what-if.xkcd.com/31

Title: FedEx Bandwidth

Found ~2 year old mail: 47.9 TB took 17.88 days (to 2x vdev of RAIDZ3, 7 disk/vdev of Seagate Exos 14 TB disks): "47.9TB stream in 1545339 seconds (32.5MB/sec)"

that's slow

I'm transferring about 250 million small files, but it's going onto a 45xhdd 12gbps enclosure

I wasn't sure how zfs send / receive would have behaved especially since I need to throttle the speed during certain hours

mbuffer can speed up zfs send | receive a lot.

yeah, isn't ssh limited @ like 125-150MB/sec due to encryption

I was going to use netcat

mbuffer can do networking too, but the primary advantage is that you can specify a sizable memory buffer that acts an intermediary for the data between the compute and transfer steps of zfs send

those are done synchronously, so without a sizable memory buffer, it takes time to move things

debdrup: thanks, I'll look into that

Greetings. I have a PC and a laptop. Both are running the same linux distribution. I have downloaded the FreeBSD-13.2-RELEASE-amd64 and two CHECKSUM files. On my laptop it says the two files(CHECKSUM and the img file itself) are "OK". But on the PC, it says "OK" only for img.xz file, and for the img file: "FAILED".

I mean, on PC-> 1) FreeBSD-13.2-RELEASE-amd64-memstick.img: "FAILED" 2) FreeBSD-13.2-RELEASE-amd64-memstick.img.xz: "OK".

on laptop-> 1) FreeBSD-13.2-RELEASE-amd64-memstick.img: "OK" 2) FreeBSD-13.2-RELEASE-amd64-memstick.img: "OK".

I do not know why, any help/advice would be appreciated.

what is saying OK or FAILED?

sha256sum command.

I'm doing checksum.

sha256sum and sha512sum.

what exact outputs do you get?

just show the sha512

Output of sha512sum? OK.

RhodiumToad: bsd.to/ok6i This is the output from my PC.

Title: dpaste/ok6i (Plain Text)

On my laptop everything is OK.

But I do not know what's wrong with my PC.

any chance you truncated the file due to lack of disk space or whatever?

what's the actual length and actual sha512 of the .img file on both?

any chance you... -> I do have enough space. I just downloaded and did the checksum on my PC.

On both systems?

yes

File sizes are the same, but the sha512 signatures are different.

if you uncompress the .xz again, is it correct or not?

What do you mean? I have already uncompressed the xz files on both systems.

you still have the xz file on both, though, going from your output?

the question is whether you get the same problem if you uncompress it a second time; e.g. is it a deterministic error, or is it some random memory or disk corruption?

Yes I still have the xz file on both systems.

the question is... -> Uncompressing again on my PC?

yes, move the .img file aside to .bad or whatever, and uncompress it again with xz -d -k ...

It said OK

:|

Why?

transient error, then

how confident are you in the condition of your hardware?

maybe compare the good and bad images to see how different they are?

I'm not even good in software :-(

How to compare?

With diff(1)?

not diff, that assumes text

RhodiumToad: One important thing

The bad image shows different signature every time I do checksum it.

Why is it like this/

uh

disk or memory problem

or some linux bug, I don't use linux myself so I wouldn't know

But computer is working fine. How can I make sure to know if something has happened to my disk or memory?

run memtest86 on it?

Do we have it on the FreeBSD?

Let me check the ports.

it's independent of OS, it's a bootable image

Thank you RhodiumToad.

I hope this time FreeBSD will be installed on my PC again.

there is a port for it, but it just downloads the image

> The bad image shows different signature every time I do checksum it.

look like hardware issue

VVD: But everything is OK. Only this SHA512/SHA256 signature was like this.

if nothing is writing to the file, and the checksum changes, then it's either a hardware problem or a very strange OS bug

and checking the hardware is probably the better first step

I should use disk health tools?

or memory health tools?

how big is the image and how much RAM do you have?

this kind of error from the actual disk isn't common, though if it's a dodgy SSD that might make it more likely

I have 4GB RAM. And the bad image size is: 1.1GB(1048712), the good one: 1.1GB(1048712)

No, I do not have SSD, it is HDD.

I'd check the RAM first.

RhodiumToad: I have no knowledge of memtest86 program. Let me test it.

RhodiumToad: OH, it is not a program. I have to boot it?

yup, it needs to get at the hardware without an OS in the way

you run it from a bootable usb drive (or cdrom or ...)

By the way, my PC is old, it uses Legacy(MBR).

that should be fine

RhodiumToad: So what's the package memtest86+ in Void Linux?

I believe it checks for corrupted memory

memtest86 and memtest86+ do the same basic things in the same way, I don't recall what the differences are

it's probably best to use memtest86plus, not memtest86

but either'll do.

Do I need a USB? I installed the memtest86+ package, and I have it under /boot directory.

they're unrelated projects, one of them (memtest86) is commercial with a free version

pr-asadi: Should be able to fire it up from Grub.

Great. Thank you.

it looks like the freebsd port of memtest86+ can be run from loader

so if you got it from linux, run it from grub; if you got it from freebsd, run it from loader

(but I don't use linux myself, so I can't help you with the details of that)

I have booted into memtest86+

It is showing a box with red color in the bottom of the screen.

saying what?

It is showing many things.

can you take a photograph

Yes. Let me.