per dire al comando che e` finita la lista di opzioni, e quel che viene dopo sono arguments

ops, E_WIN

I'm trying to setup OmniOS on Digital Ocean, it works fine, I can login and all, but I'm not able to see the console. I'm not even sure if it's a VGA screen or a serial console.

antranigv Petertribble has things hosted in DO, we could have some pointers

antranigv ptribble

neirac another interesting thing is that I cannot reboot. it shuts down, but does not come up. Should I use `reboot` or `shutdown` with some flags?

I just ssh into anything on DO, so don't recall ever looking at the console

There's an omnios mirror hosted on DO too. It's a VGA console

and the hang on reboot/shutdown is likely illumos.org/issues/14526

→ BUG 14526: illumos guest hangs on reboot under QEMU 6.0.0 (New)

We found that some of the VM types in DO work fine, but others hang like this.

andyf so I guess there's no way to bypass that right now.

Until it's fixed, wherever the problem lies, you have to stop the VM through DO's console, instead of rebooting from inside the guest. It is far from ideal.

If it's stuck, sometimes you need to get support involved, or I think somebody once found that it did fix itself after a while

and we need to work out what's going on - from what is in that issue it should be reproducible with QEMU

antranigv is your rge card still dead?

antranigv shutdown is wrapper script over setting init runlevels. reboot and poweroff commands are ok to use unless you want to have announces.

tsoome danke! learning new things every day :)

neirac going to the lab to check right now. btw I noticed that our "revisions" are not the same

andyf wish I had the knowledge to do that myself. but not really

antranigv which revision is yours?

neirac 2

neirac yours says "c"

antranigv try this rge driver, neirac.srht.site/resources/drvs/rge , you need to ipadm delete-if rge, then modunload -i (check rge id from modinfo), then copy to /kernel/drv/amd64/rge ; modload -p drv/rge , ipadm create-if rge0; ipadm create-addr -T dhcp rge0/v4 or static

hey neirac , I'm here now. I will load and try

What's different?

neirac it didn't work. dladm keeps saying unknown and I still can't send/recieve packets

any new messages in /var/adm/messages? should be debug messsages there

neirac rge_dbg_enter(0): rge_attack

neirac Using FIXED interrupt type#012

antranigv ok

rge0 registered

s/attack/attach

neirac I reset the BIOS, and now I see LINK at the other side

neirac IT'S ALIVE!

neirac what did you change?

antranigv when you say, I reset the BIOS you meant rebooted after copied the driver?

neirac yes!

Do some tests and let me know how it goes

fill a bug in illumos issues with the info you have so, if this fix is really working, it could be submitted

Well, what did you change?

interrupt handling

Just disabling MSIs?

yep

Because that's not a general purpose option.

So if that's broken, we need to understand why.

yeah, so antranigv should use hte nic to see if all is working as expected then I think we could know if that fixed the problem

Well, we don't know that the problem is so I don't think we can say it's fixed.

I agree

We've worked around and changed a bunch of things. But that driver is used for a lot of parts and I don't think we can accept a generic disable MSI fix.

oh, is only for certain models

Maybe an interim to make it more tunable to help here, but for example, the 2.5G adapters and related don't want that.

where you disable MSI

Is there a specific hardware erratum or quirk that led to that?

what do I restart to reload the routing table?

rmustacc src.illumos.org/source/xref/illumos…n/io/rge/rge_main.c?r=0dc2366f#1731

neirac should I restart the machine with the OLD module? maybe someone messed up the bios

antranigv yes, do that so we are sure the driver change was the fix

could be the bios change that made things work also

so just to be sure, restart with the old module

rebooting

ok, so this test is with the new BIOS settings and old rge module right?

yes.

works

ok, so the bios was the fix, awesome!

what was the bios change?

Load LAN ROM

I have no idea what it means

it was disabled

then I reset the BIOS settings

and it got enabled

I though hmm okay, good to know

and now it works

weird

better document that somewhere if someone has the same problem at least could search on the illumos.issues page

btw I'm getting 300kbps connection from the pkg IPS mirror. is that normal? I get ~1Gbps with our Dutch datacenter

or rather, the origin, not the mirror

who's responsible for the mirrors? I might be able to sponsor a server in Armenia

andyf could know, omnios.org/info/ipsrepos

It's the omnios core team - we have a couple of mirrors in Western US at present, and the main server is on a fast connection in Switzerland.

(At ETH Zurich)

If you configure a mirror, IPS tries to work out where it's getting the fastest responses from and adapts.

andyf how big is the mirror? I have 1Gbps, I will be getting 10Gbps this year, and we have a lot of storage that we can allocate.

They current have two disks, one 80GB for the OS (that could probably be a bit smaller), and one 300GB for the data

current size on one of the US ones is 150G on the data disk, but that varies and does go up around release time when there's an extra release running

andyf oh we can donate that much for sure. I can provide a VM for you to access, a static IP dedicated and 1TB of storage

for CPU and RAM, the host is running 256 cores and 2TB RAM. I have to allocate 1.5TB to a Ubuntu VM, but the rest is pretty much mine.

it's much smaller than I expected

antranigv that's awesome, building omnios should be quick there

yeah, you can use it for building too. I tend to dedicate the vCPUs as well. I currently allocate 200 to the ubuntu machine (altho might get pushed to 240), but the rest is there. or maybe I just keep 200 for he scientists and keep the 56 for me :P

I was happy when I had 16 cores on an old dell but the thing was too loud I was getting deaf so I had to sell it. do a test how much time takes to build omnios with 56 vcpus

In terms of RAM and CPUs, we currently have 16GB and 8 vCPUs. Let me take it to the other omnios maintainers. Thanks for the offer!

andyf thank you for keeping the sun on :)

andyf here is a current change to lxbrand that could be nice to have TritonDataCenter/illumos-joyent 80cb3c4

Yes, I saw that land in SmartOS. Nice work, we'll pull it into OmniOS next week

andyf great!, thanks Andy

Weird question. I know my way around zfs fairly well and did a lot in Solaris in the 90s. I have missing rpool capacity and can't seem to figure out why. rpool is a mirror of two ssds, 438GB in size.

rpool/build 430.12G 37.52G 409.89M 99% /build

I can't find any quotas enabled or any explanation for why it is 99%. A recursive du of all top level directories only accounts for about 60GB of the 430GB yet I'm "out of space". rpool recordsize is 128K but I can't imagine it would be so inefficient that I'd lose 370GB

The first thing I always check when people ask me about missing capacity is snapshots. Do you have any and are they taking up that space?

Only the install snapshot which isn't large

rpool/ROOT/omnios-r151048@2024-01-24-21:10:50 780M - 3.35G -

Oh cr⊙p it's swap and dump. I'm mostly a Linux guy since late 90s. Not used to boot on zfs. But thanks for pointing me in a direction

Glad to ... help? :)