#🚀 v0.15.0 - 11th December 2024

1 messages · Page 1 of 1 (latest)

atomic dirge
#

If all goes as planned, this is probably getting towards one of our last releases of the year (given we probably want to power down our release machine a bit during December due to team PTO)

This release has these changes already: https://github.com/dagger/dagger/compare/v0.14.0...main. Notably:

The milestone looks pretty clear 🌄https://github.com/dagger/dagger/milestone/69. So I'm proposing for Tuesday next week, so we got plenty of time to do another in the week if something surprising has accidentally changed.

#

cc <@&946480760016207902>

muted mauve
west tendon
west tendon
hoary bay
#

Note from prod dev weekly: we'll make this v0.15, and it'll have its own blog post, since it has all the TUI changes - and possibly some breaking changes, if they can't be avoided (OTel dependency woes + exec .Error() no longer including stdout/stderr).

atomic dirge
#

Do we need to push this out further then? If we need a whole blog post?

#

Sounds good though, we should still do what we can to avoid current modules in daggerverse breaking

#

If it's v0.15 and we have a blog post, I'd kinda of like to sneak in the new engine config as well

hoary bay
atomic dirge
#

Yeah totally arbitrary 🙂

hoary bay
#

(Thursday is Thanksgiving in the US btw)

atomic dirge
#

Hm maybe let's not go next week then?

hoary bay
#

Proposed release date: "when the blog post is ready and enough people are around" 😛

atomic dirge
#

Yeah cool 🙂

atomic dirge
#

✨ v0.15.0 - ??? 2024

#

Updated the milestone for the old v0.14.1 to be for v0.15.0 instead now 🙂

sleek aspen
atomic dirge
#

SGTM 🎉

slow sparrow
#

I have a couple of PRs on the verge of getting merged (cache volume namespacing, dagger update command and service entrypoint .

Beside that I would really love to get archive the test results from CI runs (https://github.com/dagger/dagger/pull/9011) reviewed/merged as well.

GitHub
chore(tests): archive test results in json format with CI runs by r...

this is just a poc to try out gotestsum with CI. The benefits includes a little clearer output (more work needs to be done there) and that we can archive results in a json file, which we can then i...

atomic dirge
atomic dirge
slow sparrow
#

I was waiting for the dagger uninstall to get merged, i will get dagger update ready for review (need to rebase and make sure the ci is happy)

sleek aspen
vital garnet
# sleek aspen just to be clear, I am not in a hurry to release 0.15 unless y'all are, so if we...

Given we literally just covered what's new in v0.14 in this week's community call, my preference would be to stick with the schedule we all discussed, with blogs every 2 months or so, so Mid december at the earliest or early Jan. On the other hand every rule has its exception so if we really want to release this now and there's a bunch of notable things, I'm happy to do back to back releases w/ blogs just as long as we don't start making a habit of this because I have seen that after a while people just start ignoring us if we're too noisy etc.

hoary bay
# vital garnet Given we literally just covered what's new in v0.14 in this week's community cal...

Agree with the general rule, but at this point we've already merged the big new feature and the next release will have very noticeable DX improvements and a breaking change, so it seems like we should just give it an appropriate amount of visibility/fanfare. If we're worried about saturating peoples' inboxes we could consider doing both the v3 beta Cloud launch and v0.15 release in the same blog post, we just opted not to in the interest of keeping things decoupled. (cc @meager orbit)

west tendon
#

in retrospect, do we wish we had waited on v0.14 so we could've included the v3 cloud launch and whatever breaking change?

hoary bay
#

Personally I'm more a fan of "ship whatever we have, whenever it's ready" - I wouldn't have done things any differently, since it would have meant either delaying the v0.14 release, or rushing and risking more breakage in a very visible release

meager orbit
#

Yeah I'm fine with how it went. If I remember, @hoary bay you recommended shipping the engine-side tracing improvements first, then dropping cloud v3 without an "asterisk". So the remaining question is - do we ship those engine-side improvements alongside 0.15, or 0.14.x?

#

If we feel like it's too early to announce 0.15 because we don't have enough new stuff to justify it - I guess that means 0.14.x. But since you guys are not going for that option, I guess there's a breaking change associated with the tracing improvements?

hoary bay
# meager orbit If we feel like it's too early to announce 0.15 because we don't have enough new...

There are 1-2 breaking changes:

  • Error strings no longer embed Stdout: ...\nStderr: ..., which cleans up the UI a lot but specifically breaks things like strings.Contains(err.Error(), "stdout foo"). They'll still be available as fields on the error value.
  • https://github.com/dagger/dagger/pull/8991 breaks the Bass SDK again 😢 - so anyone depending on vito/daggerverse/apko will break. Looking into work-arounds, but it's a bit of a tangled dependency knot. This is just an OTel bump and it's not needed by anything, so we can just kick the can down the road if we want.
atomic dirge
#

question - does doing all this new telemetry stuff affect the ability of old clis to talk to new engines and vice-versa?
if there's a chance of it breaking, we should probably bump the min versions here

hoary bay
atomic dirge
atomic dirge
#

<@&946480760016207902> hallo, if you have a moment today - can you add your in-flight PRs + issues to the milestone if they should be part of the release 🎉
we should try and decide now what's going in so we can work on the release blog + docs if neccessary

#

(help on clearing the review backlog would also be really appreciated!)

#

in terms of when we do the actual v0.15.0 release - should we aim for wednesday the 4th?

that gives us time if it slips a bit into thursday, and then still has time for us to publish the blog post.

sleek aspen
#

Sorry for the delay since I was OOO last week. All the PRs that are tagged as .15 milestone are added with their summaries.

@tropic jasper @spare ermine @west tendon @atomic dirge @hoary bay When you have a moment tomorrow, can you review the PR summaries in the doc? I tagged you in each section that you need to approve. Once approved, we'll mold the content into a real blog post. cc @vital garnet

#

@warped obsidian please take a look in the doc to make sure we have documentation issues for each of the PRs.

west tendon
vital garnet
sleek aspen
muted mauve
warped obsidian
sharp rivet
sleek aspen
#

all new PRs are added to the doc. @muted mauve, I made yours one section since they are all related.

sleek aspen
muted mauve
sleek aspen
atomic dirge
#

<@&946480760016207902> - I'd very much like to aim to do the release this week - it feels like we're in a very good place for it.

How does tomorrow sound? The 10th of December.

#

✨ v0.15.0 - 10th December 2024

atomic dirge
#

right, we go for tomorrow 🙂

atomic dirge
meager orbit
#

Can we also make sure the current batch of shell improvements get in?

atomic dirge
#

just want to clarify their statuses - are these blockers? as in, should these push the release back if they cannot be merged in time?

  • the shell is still experimental, so i would rather not block on shell changes - we'll aim to do a v0.15.1 the week after, so this isn't going to be as large a delay as after v0.14.0.
  • 9031 feels quite fundamental - i don't know how to judge this
meager orbit
#

Yeah we shouldn't have to block on shell merges. It's just that those are already good enough to merge as is

spare ermine
#

The shell PR is good to go. The TS one, I'm on final review, testing locally. I don't expect any issue, just double checking.

atomic dirge
meager orbit
#

Speaking of auto-complete. Any chance we can get that in also?

atomic dirge
#

if we're trying to have multiple people working on the shell at a time, having some tidy ups would be really useful

atomic dirge
#

and a review from @spare ermine

#

but, if i can get a review before tomorrow, i'll timebox a rebase early tomorrow morning
the aim is to start the release at about 13:00 UTC

meager orbit
spare ermine
#

Maybe worth pushing the release 1 day to get the tab completions in. The new shell module changed several things, so the tab-completions needs adjusting.

atomic dirge
#

it's been a month since our last release now

#

we should not be blocking the release on any experimental features

meager orbit
#

Yeah if tab-completion isn't ready to merge, it can wait for 15.1

atomic dirge
#

we'll be returning to our regular weekly schedule after this

meager orbit
#

(it reassures me that we're planning a .15.1 so quickly afterwards)

atomic dirge
muted mauve
spare ermine
atomic dirge
#

(or will try less hard)

#

although it might not be too bad

#

since the auto-complete doesn't actually use any of the shell state

#

(as we discussed, since you wanted to keep autocomplete as separate logic)

meager orbit
#

btw I can't wait to see how auto-complete handles github.com/dagger/da<TAB> 😛

atomic dirge
#

lol

#

my sneak preview - it won't

meager orbit
#

yeah I don't mean this iteration. But eventually people will ask for it, just want to prepare you emotionally for this

atomic dirge
#

(have to head out now, will pick this thread up tomorrow)

muted mauve
vital garnet
#

@solomon Draft ready for your review (draft 3, top of doc)

warped obsidian
meager orbit
#

Does 0.15 make anything else faster, besides filesync and telemetry upload?

tropic jasper
atomic dirge
warped obsidian
atomic dirge
#

@warped obsidian going through docs prs now

spare ermine
#

Yeah, both are good to go.

atomic dirge
#

awesome will hit merge 🙏

atomic dirge
atomic dirge
#

going for lunch now 🙂

warped obsidian
atomic dirge
#

merge when ready 😄

warped obsidian
#

I see some CI failures but on checking they seem to be flakes related to telemetry timeout? Ignore and merge, or re-run?

atomic dirge
#

hmmm, the daggerverse pr was never opened (investigating)

atomic dirge
warped obsidian
atomic dirge
#

re-run 😄

warped obsidian
atomic dirge
#

i suspect that what's happening is that this is only triggered on pr open - but because we add to the pr in stages, then this actually never gets correctly triggered

#

cc @balmy vector ^

#

working on updating the logic to try and skip creating the pr if it already exists

atomic dirge
#

but i'm still getting "401 Bad credentials" even though this is in the upstream repo (and not in a fork)

#

DAGGER_CI_GITHUB_TOKEN still doesn't seem set

atomic dirge
#

hm, is there maybe something about secret access in protected branches

warped obsidian
atomic dirge
#

okay i'll investigate it in a bit - it's probably some weird flake, i'll take a look

atomic dirge
warped obsidian
#

I will be afk now, dinner time here - @ me if you need something, and I will check back here after some time as well

#

merged that one now, thanks for the quick approval Justin! - 2 left to go green!

woven vessel
#

I'll send over a PR with the fix

atomic dirge
#

ah

#

i have just thrown together a fix actually

#

will open a separate pr once it passes?

woven vessel
#

Looks good. I'm actually trying to understand why we added that step. THe step before, Verify & Comment already validates if the deployment was successful or not. If it was it posts a comment

#

So I'm not sure why ensure-healthy is necessary given we already did the work before that

#

cc @balmy vector

#

I think we are okay to remove it from .github/workflows/daggerverse-checks.yml

atomic dirge
#

@woven vessel ^

woven vessel
atomic dirge
#

hm, it is starting to get kind of late here, which means things are starting to feel rushed - i was hoping to be done by now.
i think instead of rushing, i'm going to vote to hold off the release into tomorrow

#

that gives us a bit more breathing room - because it's been a month since exercising the release machine, i'd rather not rush it, end up with a bunch of stuff to fix, and end up staying up really late to get it all done

#

fyi <@&946480760016207902>

#

have merged the last thing we need, so all that needs doing is tagging v0.15.0 now

spare ermine
atomic dirge
#

looking now!

#

🤔 what

image.png
#

nvm, it's a bug in the "new merge experience" in github 😛

#

merged it now

atomic dirge
#

okay, we look good to go - tagging in a few minutes

#

finished merging all the docs prs @warped obsidian 🙂

atomic dirge
#

bleh, daggerverse bump failed, helm deploy failed - tackling helm deploy first

#

i'm gonna run the publish step manually locally I think

#

going to do the same with the bump daggerverse pr

balmy vector
atomic dirge
#

i think i'm good right now! i'm really a fan of the experience of being able to run chunks of the pipeline locally

balmy vector
#

Indeed! This is what I have for the Daggerverse bump PR:

jd -n dgvs-bump 0.14.0 0.15.0
#!/usr/bin/env bash
export GITHUB_PAT="$(op read op://Private/GITHUB_X22_PAT/credential --account my.1password.com --cache)"
set -eux
dagger -m modules/daggerverse call --github-token=env:GITHUB_PAT bump-dagger-version --from=0.14.0 --to=0.15.0
atomic dirge
#

i notice it hasn't updated all the workflows? cloud previews, and magicache tests - but it looks like everything else is pretty good 😄

balmy vector
atomic dirge
balmy vector
#

Looking now

#

Taking a closer look now.

atomic dirge
#

published new docs - includes the new config pages!

atomic dirge
#

🚀 v0.15.0 - 11th December 2024

#

working through release improvements - got a little checklist to go through
but i think the main part of the release is done 🙂

#

going to go out and grab a spot of lunch, but will be back in about half an hour to an hour

balmy vector
#

I am seeing a different behaviour in Daggerverse after the upgrade to 0.15.0. Still trying to understand what is happening. The tests are hanging.

atomic dirge
#

I think the AsService point is correct, have updated the relevant places

atomic dirge
#

okay, well we get past the tests now 😄

#

building also succeeds

#

not sure how long the deploy step is meant to take to become healthy

balmy vector
#

We both pushed the fixes at the same time. Going to force push mine since they have a few extra changes.

#

OK, all my changes are now pushed. Watching this go through.

#

Happy to jump into team chat @atomic dirge so that we don't step on each other toes again.

atomic dirge
atomic dirge
#

we can merge that pr fyi, i can do all the other bits as a follow-up

balmy vector
atomic dirge
balmy vector
atomic dirge
#

yeah i'm not actually sure what daggerverse/dag does, so, if it works 🤷

balmy vector
#

Taking a closer look at that now.

#

Those fixes were actually legit. I don't think that part of the codebase is used, but the changes are important. Putting them back now.

#

Nice catch!

sharp rivet
atomic dirge
balmy vector
balmy vector
#

Can we get an extra pair of eyes on https://github.com/dagger/dagger.io/pull/4110 @lunar bloom @half apex @woven vessel ?

It all looks good to me, I just want to give us all an opportunity to take a moment and celebrate this first auto upgrade 🎉

balmy vector
west tendon
#

i think we might have a regression in the (disk io) metrics from v0.14.0 to v0.15.0, still trying to verify

#

on a dev-built-from-tagged version a curl based demo i built does not output the disk util you'd expect

#

seems plausibly related to orbstack specifically, verifying now

tropic jasper
#

Yeah on v0.15.0 I only get metrics from services (was running dagger call test specific in our repo)... I wonder if the telemetry changes broke something since the TUI displays metrics by associating them with a CallID

west tendon
#

seems plausible, i have only seen service metrics and metrics from terminal() calls

atomic dirge
#

Hm - I have also recently sometimes seen metrics appear not in high verbosity modes? Is that intentional?

west tendon
west tendon
atomic dirge
#

When would you want to be able to release a fix? I want to do a .1 next week, does that timeline work?

#

If not, we can do sooner

tropic jasper
#

Once we figure out what's wrong and have the fix merged we can decide, but I'm happy to do one anytime. Seems like the release process is easier than ever thanks to everything that's gone into it, so not a ton of overhead

atomic dirge
#

Yup! Getting a lot of use out of the release machine is probably a good way to iron out all the last bits and bugs in it as well

west tendon
#

i wanna do one with these improvements, too 🙂

#

i demand shit to shovel and water to carry lol

atomic dirge
#

Oh you're so welcome to it bestie 😂

west tendon
#

also fun hindsight-is-20-20 realization i think i might've hit this bug last week, while playing with @tropic jasper 's hackathon --dot-output, and to quote the dm "i do not have an explanation for this" (i was only getting io metrics on services in my graphviz output)

tropic jasper
#

We need to improve the tests too, didn't include that there though

west tendon
#

stamped, testing on my end now

# 
│ ✔ .withExec(args: ["bash", "-c", "for i in {1..20}; do curl -s https://w.wallhaven.cc/full/jx/wallhaven-jxzzmp.jpg -o \"cat$i.jpg\" && sync && sleep 0.1; done"]): Container! 7.6s | Disk Read: 0 B | Disk Write: 76 MB | IO Pressure: 6.649ms | CPU Pressure (some): 8.953ms | CPU Pressure (full): 8.953ms | Memory Bytes (current): 79 MB | Memory Bytes (peak): 82 MB | Network Rx: 605 kB | Network Tx: 100 MB

working on colima at least, orbstack weirdness is likely unavoidable

atomic dirge