imagine someone reading the commits later and that shows up. ahhahaha at least it shows the history

RRobert How does this work? I couldn't see any podman push or artifact upload steps aft...

B

bsherman•2/2/25, 8:46 PM

oh, i know what you mean... why don't you see the kernels fetched? because an earlier build successfully cached them already

Ttulip imagine someone reading the commits later and that shows up. ahhahaha at least i...

N

Niklas ⚡•2/2/25, 8:46 PM

you should use more sobs in your ones

Bbsherman oh, i know what you mean... why don't you see the kernels fetched? because an ea...

R

Robert•2/2/25, 8:46 PM

No - I'm meaning you're calling the cache action before making the kernels. But I guess the cache action saves the data in a post-workflow step?

RRobert No - I'm meaning you're calling the cache action _before_ making the kernels. B...

B

bsherman•2/2/25, 8:47 PM

yes, that's how it works by default.

there's a way you can be more specific and "retrieve from cache" and "save to cache" but this seems fine

M

M2•2/2/25, 8:48 PM

this looks like it is nontrivial to build locally now...

RRobert No - I'm meaning you're calling the cache action _before_ making the kernels. B...

B

bsherman•2/2/25, 8:48 PM

here's an example of that post-workflow step NOT saving due to cache-hit https://github.com/ublue-os/akmods/actions/runs/13102055564/job/36551528729#step:17:2

MM2 this looks like it is nontrivial to build locally now...

B

bsherman•2/2/25, 8:48 PM

that IS the downside

M

M2•2/2/25, 8:49 PM

Probably can throw together a Justfile as part of the testing

MM2 Probably can throw together a Justfile as part of the testing

B

bsherman•2/2/25, 8:49 PM

yeah, i do something like this already at work, and i use just file to prep a local cache which matches up like this

B

bsherman•2/2/25, 8:50 PM

I'd like to convert this to Just and clean it up a lot, but I felt it more important to move forward iteratively

B

bsherman•2/2/25, 8:50 PM

less overall change makes it easier to grok the set of changes... I hope.

M

M2•2/2/25, 8:50 PM

yeah i think being tied first to github actions is fine given that we are relying on the cache action

B

bsherman•2/2/25, 8:51 PM

@p5 here's where the cache was last saved: https://github.com/ublue-os/akmods/actions/runs/13101891946/job/36551158340#step:17:1

MM2 yeah i think being tied first to github actions is fine given that we are relyin...

B

bsherman•2/2/25, 8:52 PM

right, but we could still move most of the bits into Just recipes for a local build option and then change the workflows to call those same recipes

again, I consider that future improvement which would further delay progress

M

M2•2/2/25, 8:53 PM

Secureboot check only happens if no hit on the kernel. That's smart

B

bsherman•2/2/25, 8:55 PM

yeah, the cache-hit check feels quite powerful, if we are confident in the cache, we can speed things up but also be confident we're re-using the same artifacts

M

M2•2/2/25, 8:56 PM

we should probably remove the --mount=type=cache,dst=/var/cache/dnf from the containerfiles

R

Robert•2/2/25, 8:56 PM

How close are we to the 10GB cache limit in that repo? I have no idea how large kernels are

M

M2•2/2/25, 8:57 PM

the current cached rpms are usually about 200 MBs

M

M2•2/2/25, 8:57 PM

Surface is slightly larger but we could probably remove the additional cached packages from them

J

j0rge•2/2/25, 8:58 PM

@ssweeny we ever migrate that repo over?

Jj0rge @ssweeny we ever migrate that repo over?

S

ssweeny•2/2/25, 8:59 PM

Both of my akmods repos are migrated

J

j0rge•2/2/25, 9:00 PM

ok removing your temp org membership, thank you for flying universal blue airlines

M

M2•2/2/25, 9:00 PM

https://github.com/actions/cache?tab=readme-ov-file#cache-limits

So it's 10 GB and gets evicted after not being used for a week.

The kernel rpms end up in each container?

B

bsherman•2/2/25, 9:03 PM

anduril in akmods on  combine-kernel-kmods [»!?] took 3s 
$ gh cache list

Showing 6 of 6 caches in ublue-os/akmods

ID  KEY                                      SIZE        CREATED           ACCESSED           
17  surface-6.12.7-1.surface.fc41.x86_64     113.97 MiB  about 1 hour ago  about 5 minutes ago
18  asus-6.12.9-202.rog.fc41.x86_64          192.53 MiB  about 1 hour ago  about 5 minutes ago
13  coreos-stable-6.12.9-200.fc41.x86_64     192.38 MiB  about 1 hour ago  about 5 minutes ago
15  main-6.12.11-200.fc41.x86_64             192.37 MiB  about 1 hour ago  about 5 minutes ago
16  bazzite-6.12.11-206.bazzite.fc41.x86_64  194.62 MiB  about 1 hour ago  about 5 minutes ago
14  coreos-testing-6.12.10-200.fc41.x86_64   192.35 MiB  about 1 hour ago  about 5 minutes ago

anduril in akmods on  combine-kernel-kmods [»!?] took 3s 
$ gh cache list

Showing 6 of 6 caches in ublue-os/akmods

ID  KEY                                      SIZE        CREATED           ACCESSED           
17  surface-6.12.7-1.surface.fc41.x86_64     113.97 MiB  about 1 hour ago  about 5 minutes ago
18  asus-6.12.9-202.rog.fc41.x86_64          192.53 MiB  about 1 hour ago  about 5 minutes ago
13  coreos-stable-6.12.9-200.fc41.x86_64     192.38 MiB  about 1 hour ago  about 5 minutes ago
15  main-6.12.11-200.fc41.x86_64             192.37 MiB  about 1 hour ago  about 5 minutes ago
16  bazzite-6.12.11-206.bazzite.fc41.x86_64  194.62 MiB  about 1 hour ago  about 5 minutes ago
14  coreos-testing-6.12.10-200.fc41.x86_64   192.35 MiB  about 1 hour ago  about 5 minutes ago

anduril in akmods on  combine-kernel-kmods [»!?] took 3s 
$ gh cache list

Showing 6 of 6 caches in ublue-os/akmods

ID  KEY                                      SIZE        CREATED           ACCESSED           
17  surface-6.12.7-1.surface.fc41.x86_64     113.97 MiB  about 1 hour ago  about 5 minutes ago
18  asus-6.12.9-202.rog.fc41.x86_64          192.53 MiB  about 1 hour ago  about 5 minutes ago
13  coreos-stable-6.12.9-200.fc41.x86_64     192.38 MiB  about 1 hour ago  about 5 minutes ago
15  main-6.12.11-200.fc41.x86_64             192.37 MiB  about 1 hour ago  about 5 minutes ago
16  bazzite-6.12.11-206.bazzite.fc41.x86_64  194.62 MiB  about 1 hour ago  about 5 minutes ago
14  coreos-testing-6.12.10-200.fc41.x86_64   192.35 MiB  about 1 hour ago  about 5 minutes ago

anduril in akmods on  combine-kernel-kmods [»!?] took 3s 
$ gh cache list

Showing 6 of 6 caches in ublue-os/akmods

ID  KEY                                      SIZE        CREATED           ACCESSED           
17  surface-6.12.7-1.surface.fc41.x86_64     113.97 MiB  about 1 hour ago  about 5 minutes ago
18  asus-6.12.9-202.rog.fc41.x86_64          192.53 MiB  about 1 hour ago  about 5 minutes ago
13  coreos-stable-6.12.9-200.fc41.x86_64     192.38 MiB  about 1 hour ago  about 5 minutes ago
15  main-6.12.11-200.fc41.x86_64             192.37 MiB  about 1 hour ago  about 5 minutes ago
16  bazzite-6.12.11-206.bazzite.fc41.x86_64  194.62 MiB  about 1 hour ago  about 5 minutes ago
14  coreos-testing-6.12.10-200.fc41.x86_64   192.35 MiB  about 1 hour ago  about 5 minutes ago

MM2 https://github.com/actions/cache?tab=readme-ov-file#cache-limits So it's 10 GB ...

B

bsherman•2/2/25, 9:04 PM

read the docs

yes all kernel RPMs end up in each akmods image at

/kernel-rpms/

/kernel-rpms/

/kernel-rpms/

/kernel-rpms/

Bbsherman ``` anduril in akmods on  combine-kernel-kmods [»!?] took 3s $ gh cache list ...

A

antheas•2/2/25, 9:14 PM

what are you caching?

Aantheas what are you caching?

B

bsherman•2/2/25, 9:41 PM

feel free to take a look and provide feedback: 💾ublue-dev

Bbsherman feel free to take a look and provide feedback: https://discord.com/channels/1072...

A

antheas•2/2/25, 9:42 PM

why do we cache the kernel?

A

antheas•2/2/25, 9:43 PM

10gb means 10 kernels?

A

antheas•2/2/25, 9:43 PM

is that right?

A

antheas•2/2/25, 9:44 PM

no we have 6 kernels

A

antheas•2/2/25, 9:44 PM

ballpark 8 kernels?

A

antheas•2/2/25, 9:44 PM

can the kernel cache be used to rollback?

Aantheas can the kernel cache be used to rollback?

B

bsherman•2/2/25, 9:46 PM

I think the question would be answered by looking at the PR.

But the github kernel cache is an internal implementation detail for builds.

The only end state change is to include the Kernel RPMs (currently in

kernel-cache

kernel-cache

kernel-cache

kernel-cache images) in the

akmods

akmods

images in the

/kernel-rpms/

/kernel-rpms/

/kernel-rpms/

/kernel-rpms/ directory.

A

antheas•2/2/25, 9:47 PM

im looking at it and cant tell

A

antheas•2/2/25, 9:47 PM

i tend to no

A

antheas•2/2/25, 9:47 PM

so, whats the point of the cache?

A

antheas•2/2/25, 9:48 PM

i like the grouping

A

antheas•2/2/25, 9:49 PM

wow the cache takes 2 seconds?

M

M2•2/2/25, 9:49 PM

Don't download the RPMs everyday and make sure all of the packages are built from same kernel version

M

M2•2/2/25, 9:49 PM

2 seconds vs 4 minutes

MM2 2 seconds vs 4 minutes

A

antheas•2/2/25, 9:50 PM

im seeing 20 seconds?

A

antheas•2/2/25, 9:51 PM

in any case, implementation detail no rollbacks

A

antheas•2/2/25, 9:51 PM

can you make an action that takes a bazzite kernel version, defaulting to latest, and builds that kernel?

A

antheas•2/2/25, 9:51 PM

we can also webhook that

Aantheas im seeing 20 seconds?

B

bsherman•2/2/25, 9:55 PM

the best case cache hit is a bit faster/slower depending on method used to lookup latest available kernel version

B

bsherman•2/2/25, 9:55 PM

it varies for kernel flavors

A

antheas•2/2/25, 9:55 PM

yeah bazzite kernel is instant

Aantheas can you make an action that takes a bazzite kernel version, defaulting to latest...

B

bsherman•2/2/25, 9:55 PM

i'm sure we can, that's not in scope for the current PR

imagine someone reading the commits later and that shows up. ahhahaha at least it shows the history

Similar Threads

Similar Threads

Similar Threads