hi, im new to smartos. have everything installed but when i run imgadm import uuid, i get disk is full

not sure where it's trying to save, but the lowest capacity is zones/cores which has 10gb free.

and / is a ramdisk that's practically full

... and /zones has about 15gb free

15GB isn't all that much space for a useful host, but all the image I've ever looked at are sub GB in size.

it's just for some testing purposes, not for being a production server

regardless, it won't let me import anything because / is full

is always fairly full.

Maybe /tmp?

How much ram?

about 20gb ram.

not sure about tmp, i didn't look, it was complaining about /

oh, it specifically said / is full?

yup

it downloaded like 16mb and then exited with that error

so what does `df -k /` look like?

It should be doing any downloading into /var/tmp.

sorry for the delay. df -k / is 95% full

i dont see any mount for /var or /var/tmp, so i guess these go into /, but /tmp is 1% full

sorry, /var is 1% full

both /var and /tmp have about 15gb free space.

pjustice ^

Would be more useful to know bytes than % for /

(I'm basically comparing to one of my host machines here.)

17165 free

303637 used

Oddness.

I think at this point, I would try running imgadm under `truss` to find out what files it's creating. Having a second terminal session open to look in e.g. /tmp might be useful too.

i mean, i run this from a livecd (i think that's like the only option?), so im not sure how did i even end up in this mess to begin with

i don't mind re-starting the whole thing either

it has literally nothing in there so far

You can boot it from a usb stick or pxe boot it too. And once you get a machine installed, you can actually set it up to boot from local hard disk.

I've never actually run a machine with that little ram and disk, so I don't know where it gets into trouble.

well it's running from a live cd right now

i guess thats really a usb stick

Shouldn't really matter.

i've literally just "installed" it and configured a zpool. that's all i've done.

and installed probably means configured it, not even installed anything

Yeah, you have to do something else I never bothered to look up to get it to make the local disk bootable.

Given the numbers you're seeing, and what the source code says imgadm does, this really isn't making sense to me. Several of the developers are around during western hemisphere daylight hours. And I'd try the `truss` idea if you haven't yet.

danmcd: rmustacchi: I took shipment of 50 Intel E810-CQDA1 NIC's yesterday...against my will lol. Do either of you happen to have the bug/feature request number for the illumos driver? Also, I think I asked this before but I cannot remember...(Can also test tonight)...will the 100GBe NIC link up to a switch at 40GBe with the i40e driver?

I can't figure out what happens to my nick nowadays. I was logged in as barfield but it changed my name to Guest1332 after I joined #smartos. Something must be going on with m~y client.

bahamat: I noticed the packer images for ubuntu 22.04 released in November 23 have focal fossy or whatever the name is for 20.04 in the description. Just thought that I would report it.

Instead of "Jammy Jealyfish".

Ok, I'll take a look.

One other thing; is github the right place to open issues? Or should they just be reported on the mailing lists/irc so that it goes to atlassian?

You can report issues with github.

uf: You might be getting that error from bash not being able to write the .bash_history file. But imgadm isn't going to write to /, images go in /zones.

uf: What does `zpool list` say under FREE?

Guest1332: I don't know if thre's an open bug there. I did a small start of this in 2019 while still at Joyent that had some basics, but no I/O. github.com/rmustacc/illumos-gate/tree/ice-dev.

But there is no complete illumos driver for that part. The E810 will not work with the i40e driver.

Thank you sir!

My partner should be thrilled :)

At least I get to say I told yo so

It's a good chance to learn device development in my opinion.

Nothing is motivating like a lot of parts you need to work!

Guest1332: triton-cloud-images#14

There's a scaffold of it so far and importantly the I/O engine is the same as i40e.

So for better or worse you can leverage that.

Would sponsorship be of any interest to you or any other developers to reprioritize this work?

Sorry if that breaks any IRC rules

thought about the rules after I hit enter

I'm not in a position where I can. Others might.

We don't have any particular rules about it, but people need to prioritize their day jobs.

Thanks I definitely agree

The best way to influence what gets worked on is by having a support contract where you can make specific requests. Then it's up to the business to set the appropriate priority.

We've got some things in the works coming up that Nick/MNX may be interested in so I'll reach out in a week or so.

I can answer questions and other things, but this is not an effort I can lead and not really a spare time kind of thing for me tbh.

Your answers are valueable. Thank you

I tried to implement the s3-manta-bridge recently...and I could see the API requests working correctly but data was never visible in buckets from s3cmd. Does anyone here know if all of the s3 API projects interfacing with manta were broken by the v2 release?

I wouldn't describe it that way.

mantav2 has two interfaces for storing objects.

1. directory style, the API is unchanged from mantav1 and this is still the only one officially supported

2. bucket style, completely different semantics and internal indexing. This is experimental only.

Thats the buckets-api right?

Yes

Experimental is not officially supported but if I recall while still at Joyent, it was working right?

If you use the buckets branch of node-manta you get the mbucket command, which can be used to copy objects in/out (i.e., PUT/GET)

interesting

It pretty much works. But there are still things with it that preclude declaring it as GA and fully supported.

As of now, if you use it I can't recommend putting production workloads on it.

We don't really know how well it scales yet.

Both as in what happens when it gets really full, and what happens when you need to grow it.

The mbucket command should work.

So if you're trying to make manta buckets be the back end for your s3 front end, then you'll first need to make sure that the equivalent mbucket command is working properly.

Assuming it works, then you need to check the difference in the API calls that your thing is making.

thanks!