vortexx: sorry little bit ago you stated to zpool import the zvol, this does not work on zvol's it appears.

jaredj: I have tried mdconfig -a on the zvol but I can't seem to figure out how to get it to work, afaik this mounts it as a memdisk, however I get "mdconfig: /dev/zvol/zpool-storage/dns is not a regular file" so I am clueless on what to do

polarian: not sure what you are trying to do, but zvol is already a block device, you shouldn't need mdconfig

except for setting proper 'volmode', most likely

actually no, reading zfsprops(7) default is already "full", giving you fully functional block device.

yuripv: the zvol has a partition table; polarian needs to grok that, obtain block devices for the partitions, and mount them

jaredj: still that doesn't need mdconfig, as above; if it was used as a "disk" by bhyve, it probably has 'volmode' set to 'dev' and need it set to 'full' and then check `gpart show` output (e.g. dpaste.org/6b5bz)

yuripv: gpart: No such geom: /dev/zvol/zpool-storage/dns.

you cant interact with the block device

specify it as zvol/zpool-storage/dns

oh

polarian: sure you can

yuripv: did that, same error

gpart: No such geom: zvol/zpool-storage/dns.

but first, what is volmode for that zvol?

zfs get volmode zpool-storage/dns

yuripv: dev

that's the problem, set it to 'full'

if you are NOT using it as raw device at the moment, that is

yuripv: oh right

is this mentioned in the man pages

the volmode? zfsprops(7)

volmode

I have searched, nothing is found

would like the reference material not just spoon feeding (no offence) :)

oh sorry I was in zfs(8) not zfsprops, thanks :)

yeah, properties were moved to separate man page from zfs(8)

ok that fixes it \o/

thanks yuripv

yw

hmmm, bsdlabel/disklabel is deprecated (use gpart) but I can't see any reference to it in the man page... gpart show -p zvol/zpool-storage/dns shows zvol/zpool-storage/dnss4 (which is disklabeled) but "geom doesn't exist"

I am new to zfs if it isn't obvious :P

that's not about zfs already

I know the other stuff was

zfs part is over :D

this is me not knowing gpart very well :P

I rarely need to use it so its one of those forgetable things :P

what are you trying to do with those partitions? mount?

oh wait I dont need them partitions they contain the data, the first partition contains the kernel iirc

hmm mounting the partition doesn't work "file not found"

oh wait its zfs... you set it to online iirc

wait no...

I should just re-read the handbook and man page again sorry for the noise

yuripv, thanks for correcting my incorrect advice

(its zfs-import) :P

zpool*

yuripv: is it even possible to mount a zvol... zfs won't touch it because "operation not applicable to datasets of this type"

and I can't find any documentation on mounting zvols so I am wondering if its even possible...

zvol is block device not file system

tsoome_: exactly

but it contains a filesystem...

a filesystem I need to inspect

tl;dr it contains an OpenBSD installation being used by bhyve, but /bsd can't be found

polarian: zpool(8) deals with the block devices underpinning filesystems, zfs(8) only deals with the filesystems on top, iiuic

no matter what, I can't boot from the zvol despite it being a fresh OpenBSD install... I have been trying for months

and then putting it off for another month...

zfs set volmode=geom <the-zvol> ?

others who followed my steps say its fully working, they can boot just fine... I can't for the life of me get it working, so I am trying to inspect the filesystem on the zvol to see if there is any noticable reason

jaredj: geom == full, its in full mode right now

aw.

zfsprops(7): "The value geom is just an alias for full and is kept for compatibility."

so i see after rtfming

so do you see it in geom disk list?

no

ooo, geom -t, didn't know about that one

I have mounted FFS2 partitions using the UFS driver on FreeBSD before... but I can't find any way to mount the FF2 partition(s) from a zvol to the host

that is what I am trying to figure out... and I feel like I am being really stupid and missing something obvious

well - the problem is the FFS2 is inside a partition on the zvol, right?

so you can't just mount the zvol

you need per-partition block devices to show up somehow

"...Setting [volmode] to dev hides its partitions." wat

jaredj: indeed...

but gpart does show, (as suggested by yuripv)

gpart knows there are partitions? but then you do not get devices under /dev/zvol for them?

jaredj: exactly :)

well actually technically I do

but the partitions under /dev/zvol can't be mounted with mount

and the filesystem isn't zfs

what does file -s say about those partition devices

/dev/zvol/zpool-storage/dns: DOS/MBR boot sector; partition 4 : ID=0xa6, active, start-CHS (0x0,1,2), end-CHS (0x3ff,254,63), startsector 64, 41942976 sectors

so the block device is exposed as a filesystem

uh, but you don't have /dev/zvol/zpool-storage/dnsp1 for example?

have you tried switching back and forth between volmode=dev and volmode=full?

jaredj: no

there is slices there is no partition labels

so I have dns and dnss4

oo, what's file -s say about dnss4?

s4 will contain disk label for the entire of the OpenBSD filesystem

/dev/zvol/zpool-storage/dnss4: DOS/MBR boot sector, code offset 0x3c+2, OEM-ID "OpenBSD", sectors/cluster 2, Media descriptor 0xf8, sectors/track 0, hidden sectors 16, serial number 0, label: " UNIX LABEL", FAT (1Y bit by

descriptor); NTFS, physical drive 0x290000, sectors 2330693407946571776, $MFT start cluster 5067957283254845772, $MFTMirror start cluster 5254012139394572371, clusters/RecordSegment -1341669376, serial number 0e7bfcbcd08ec031, checksum 0x1d8be1f

uhrrr, right, that two-level thingy

i just installed openbsd on an old laptop recently and ... i don't like bsd disk labels

:)

bsdlabel won't read it though (plus its deprecated)

right so you need /dev/zvol/zpool-storage/dnss4a to somehow exist

"Use gpart(8) instead (with the BSD partitioning scheme)," ok

i lost track - does gpart list say anything more detailed than s4?

jaredj: no

Does anybody use syncthing? How do you deal with high cpu usage?

40k files, 2500 dirs, 120G, and cpu is not really doing anything when there's no file sync happening... haven't found cpu to be a problem ever though s2r

polarian: as far as i've read, having geom deal with a bsd disklabel inside an MBR partition table is utterly usual and indeed an example of the sort of generality for which geom was built

it's just that all the examples are with disks not zvols.

can you gpart show /dev/zvol/zpool-storage/dnss4 ?

s2r: i use syncthing, i've not noticed much cpu

but i only have it check every hour or so

After upgrading to 13.3-RELEASE-p5, it seems sslh keeps crashing thereby preventing me from ssh'ing to the server. I've been using sslh for years and have not face such an issue. The logs don't seem to have anything in it. Where would I find a core dump of sslh, if there was one?

does it happen every time?

i would run sshd -Dddd and then connect wish ssh

it doesn't happen with every ssh connection.

Last time I restarted it was on the 22nd of August. I've had several SSH sessions since then. I was able to SSH in earlier today before 1700 UTC. I went to ssh in again around 2030 UTC and I was not able to connect to sslh. sshd is up and running just fine.

s2r: I do not use synthing (yet... it is the prefered way to backup android phones), however I have not heard anyone I know who uses it complain about high CPU use... you sure there isn't another issue happening?

jaredj: the issue I am having is for my entire time on *nix systems I am used to dump filesystems (FFS2, UFS and ext4) having such an advanced filesystem causes a lot of confusion... for example you need to ensure the zpool is attached, then the partitions (which aren't really partitions as they are virtual not physical) can be mounted but not with mount but with zfs-mount... its just a whole new

can of worms...

I do like the compression and the checksumming... but its very advanced

sslh?

sslh? i thought that was a typo

rtprio: yes SSLH, not SSH :-) Not a typo. SSLH multiplexes incoming connections, a simpler HA-Proxy of sorts.

mns, Looking at sslh it will run as user nobody in / and therefore will not her permission to write a core file there. Modify the /usr/local/etc/rc.d/sslh file so that it runs in a directory that it can write to and see if it dumps core to that now writable directory.

*will not have permissions

Does sslh also crash when it the other protocols it is multiplexing are used to pass through it?

rtprio, The problem sslh solves is when I take my laptop to the local hospital here to visit a friend and then try to use the laptop on their network they have a very restrictive firewall that blocks everything except for HTTP/HTTPS port 80/443 and nothing else is allowed through blocking my ability to ssh anywhere. But port 443 is open.

I always keep a server with sshd running on port 443 that I can connect into for this reason. Then I can ssh out of this restrictive system. But then that server can't run a web server which also requires port 443. Enter sslh which will sit on port 443 multiplexing the connections to the web server and to sshd as required to make sharing of that port possible.

I also notice that the binary pkg port of sslh has a zero sized sslh.8 man page for it. Error in the port.

polarian: and the snapshots! ... i guess you can have those on dump filesystems, hm

I dont get the attraction to sshl

a VPN would give more flexibility

you will have to obfuscate and mask the packets to prevent being dropped from deep packet inspection

jaredj: yes snapshots are cool too (makes backups easier)

snapshot and then use send/recv to distribute it to many other pools

yeah. i was just trying to play around with snapshots, clones, nullfs, and -o union to make a filesystem with a read-only base and a read-write overlay. it didn't work though