when downloads.omnios.org/media/bloody gets rebuild? my miniroot just fails on pci symbols, so I thought I better wait for the ones in here

That sounds like you're missing omniosorg/kayak fadaf6e which is a couple of months old. Are your repos all up to date? There will be a new bloody release and media sometime today, which will have the vioscsi driver.

Bloody is usually updated weekly, and media every few weeks depending on what's changing

Hmm latest r151046 seems to now hang crash on linode with imgur.com/nwtEHS5

andyf it's only a 1G memory system though, so maybe one of the recent changes is causing issues but I would not expect hat withint a release to happen

interesting, booting an older BE doesn't help either

the smb stuff in the stack are weird though

I haven't been able getting a succesful boot

while I have 2 other VMs running on the n release

Anybody know an illumos friendly EU hosting provider that does not cost an arm and a leg per month?

Linode was great for the last 3ish years but seems there new hypervisors don't boot illumos anymore

Looks like I am now hitting github.com/illumos/illumos-gate/blo…common/vm/vm_page.c#L2311C1-L2312C1

Which hasn't been changed for 19 years

That would also explain why I am hitting it on all versions I have going back

imgur.com/nwtEHS5

Looks like it got booted on a newer hypervisor

I've had some success with Digital Ocean

Tribblix is currently hosted on AWS

The costs for the instance and storage are actually the same, but AWS charges a fortune for bandwidth whereas DO gives you a good inclusive allowance

I never got omniops to work on DO when I tried

had a look at hetzner as some collegues use that for there linu stuff and they seem happy

But can't get omnios to boot either

Well I tried dd the cloud.raw and bhyve images via the resqueue system

But no luck

Ideally I'd like to get linode to boot again though, pretty happy with them

It's probably the disk type on hetzner tbh

I'll check what it is in the rescue image

Yeah vioscsi

sjorge - that seems unfortunate, the new hyperisor.

Yeah, luckily my mail server is still up that DC is schedule to be update next months

I wonder what they've done that triggers this for you (although 1G is really minimal with ZFS)

some maintanance

in the Dc it was hosted

So I assume they rebooted the old hypervisor and it brought the VM up on a newer node

1G has been a bit tight but but working pretty well for my tiny work loads

I looked around but there is no tunable to it seems to try and change the bahvior to try and allocate less

At least not that I found

andyf do is there a raw image somewhere that has the vioscsi driver installed (and boots over graphics console)

No serial on hetzner

but there is a VNC based kvm

No. The next raw image will have vioscsi, but all of them use the serial console.

I think you can get to loader early via VGA and reconfigure it though

I only saw the BTX line but nothing after that one... one of the cloiud images

Do the bloody images have vioscsi ?

I could grab one of those too I guess

The bloody cloud images have vioscsi, yes, the installer does not yet.

but you don't need the installer if you use cloud images

OK let me try with that one

it might not show more than the BTX line because loader can't load it's confiog

Let's see what omnios-bloody-20240215.cloud.raw.zst does

imgur.com/ibP5DuB same

Just this and nothing

andyf do you think the 1G is the issue, I could bump it to 2G for 2x the price I pay now

But I can never shrink it if that was not the problem

Yes, I think it probably is. It seems pretty clear that it can't allocate enough kernel memory. What would be interesting is why it either need to allocate more than it used to, or why there is less available now. Either could be true based on the hypervisor or SMBIOS data or whatever.

Let's try that I guess

It looks to me that we're getting nonsense out of smbios

Worse case it's one month of extra billing as if it doesn't work I'll probably delete it once I do find a place to run.

Can you exclude the smbios kernel module?

It's still resizing to 2G linode vm, can take up to 1h :|

But I do have loader access so I think I could probably disable that

Dropping to the 'ok ' prompt and typing 'boot -B disable-smbios=true' is worth a try

gist.github.com/sjorge/e289838988ce141096cdb4d9a6242e34 2G image did not help, dif error now

will try with boot -B disable-smbios=true

OK now I can't catch the loader prompt as it's gone to fast -_-

probably because this image also adds an extravcpu

Same with smbios disable

sjorge: spearhead.cloud is a Triton public cloud provider in Romania. I'm not sure about pricing, but at least they are illumos friendly!

Interesting, adding it on the list of to check out

If you need an image that drops to the debugger by default, I can build one.

Ooh pricing is nearly 4x of what I pay now on linunode for spearhead

That's a bit much

At that point phrotonmail + a tiny freebsd VM makes more sense

@andyf I'm booting from disk on linode

But when I can beat the boot to loader, I tried kmdb

But it never get's there

It hits the issues super early in boot I think

Hmmm seems proton mail is cheaper than I though, I'll have to seriously consider that option

It would save me some pain

You can try -B kbm_debug=1 for more early output too.

Seems i can get freebsd to install on hetzner... I guess I'll rebuild on that so my site is back up today and then figure out if the linode one is salvagble

Hey all; we did the following: zfs snap -r rpool/zones/zone0@ready; zfs send rpool/zones/zone0@ready | zfs recv rpool/zones/zone1 ; zonecfg -z zone0 export -f zone1.cfg ; (modify zone1.cfg) ; zonecfg -z zone1 -f zone1.cfg; zoneadm -z zone1 attach -F;

at this point we can boot zone1, however, the IP address is not set...

in zone0 there was zone0/_a address. I assumed there would be also zone1/_a in zone1, but there's not

any thoughts?

okay... we just needed to change the IP interface name in /etc/ipadm/ipadm.conf in zone1

antranigv ah the cloning the zone took all the data to the new zone but it did not repect the allowed-address as ipadm.conf had 'invalid' config?

neirac yes. just needed to sed the ipadm.conf file in the new zone.

I don't understand why the zone with allowed-address needs an ipadm.conf file anyway.

So the two concepts are kind of independent.

allowed-addresses is about saying from outside the zone what's allowed or not to come from that zone's netstack.

sjorge, nwilkens: I had bad experiences with Spearhead staff. and I tried twice, few years between tries, thinking that maybe it was me...

It has nothing to do with what addresses are used inside the zone.

In general manually editing ipadm.conf is fraught and not intended to be used with sed.

But basically that is a record of what IP interfaces and properties should exist.

So that's always something that exists on a per-zone basis as the two ipmgmtd instances are independent effectively (it's a bit more nuanced, but it's better to think of it that way).

If you think of the zone as an independent host logically, it makes sense that it'd have to have its own networking configuration.

The allowed-addresses really is just a way of providing an ACL of what addresses the zone can send traffic out with.

rmustacc got it

rmustacc I still have to think about the best way to do a deep copy "zfs send/recv" of a zone, either on same host or another

in our case we wanted to create ldap1 from ldap0, where ldap0 is the current prod, but ldap1 should be our testing.

Another question raised while we were testing things... zadm snapshot/rollback

We got an error saying "WARNING: rollback is not supported on BE-aware brands."

So I thought hey, I'll use beadm from inside the zone... I guess?

I need to figure that out

basically I need to zadm shutdown ldap1, rollback to a be named ready, and then continue

continue == zadm boot ldap1; destroy things again by testing LDAP :P

antranigv: what about: zfs snapshot of filesystem ldap0, then zfs clone to ldap1 and adjurst paths eventually. costs are near zero.

tomww as a matter of fact, the `zoneadm -z ldap1 clone ldap0` command does that automatically and work flawlessly. I was worried that we might "promote" ldap1 to production if this large changed worked fine, so I really wanted a deep copy.

this prtconf -D output means the driver is attached or not ? pci108e,8 (driver name: vioscsi)

because I see also pci108e,1, instance #0 (driver name: vioif) , and vioif I do see in modinfo output

That generally means it's attached.

Good thing most of my ‘can get it to work on illumos’ VMs are home are freebsd, website back but on freebsd

Should be cheaper too starting next month

Now i need to figure out ehat to do for email before thst vm bites the dust

sjorge what website is this? :)

Just my personal ones

A rarely used blog :|, travel pictures, my map software I use for my dnd games, ...

prtconf -d says this pci108e,8 (pci1af4,1004) [Red Hat, Inc. Virtio SCSI] (driver not attached) ; I don't see any message about why the driver did not attached