Nightly downlod page stuck at hrev55848

[khallebal]

The page hasn't seen updates since 10 days

[kallisti5]

indeed...

`/tmp/build/c5023478/images.x86_64/haiku-master-hrev55877-x86_64-anyboot.zip` -> `remote/haiku-nightly/x86_64/haiku-master-hrev55877-x86_64-anyboot.zip`
mc: <ERROR> Failed to copy `/tmp/build/c5023478/images.x86_64/haiku-master-hrev55877-x86_64-anyboot.zip`. The bucket you are attempting to access must be addressed using the specified endpoint. Please send all future requests to this endpoint.
Total: 0 B, Transferred: 0 B, Speed: 0 B/s
https://s3.wasabisys.com/haiku-nightly/x86_64/haiku-master-hrev55877-x86_64-anyboot.zip was successfully uploaded!

Investigating...

[kallisti5]

so.. wasabi changed how their systems work. Previously we always connected to s3.wasabisys.com for every bucket. Turns out the nightly bucket (early) is in a different region than everything else.

Wasabi now requires different s3 endpoints for different regions (which is fair and makes sense tbh). The issue is our build system was written assuming a single region for artifacts.

The solution is i'm going to migrate the old haiku-nightly bucket from us-west-1 to us-east-1 (which matches everything else).

There will be a minor duplicate charge, ($8 + $8 instead of $8) for a few months... but honestly given how cheap wasabi is, it's an easy call.

I did cull some of the nightly images using nightly-reaper-s3.rb since they were getting a bit out of control...

Before:

• 917 GiB 8217 objects haiku-nightly

After:

• 498 GiB 5668 objects haiku-nightly

I'm beginning the mirror of haiku-nightly (us-west-1) to haiku-nightly-us (us-east-1) now. It'll take several hours.

[kallisti5]

ok. i've migrated over to the new bucket (haiku-nightly-us) and erased the old one (haiku-nightly)

I'm running a toolchain build to get efi signing tools and git-restore-mtime into the toolchain containers. Then i'll run a build of of the nightly to see how it goes.

[kallisti5]

resolved. ​https://ci.haiku-os.org/teams/nightly/pipelines/master-x86_64/jobs/image-master-x86_64/builds/912

We'll see new nightly images tonight.

Thanks for reporting this one!

[rudolfc]

Hi, looks like 32bit is still down.. (or am I too soon with this message?)

[madmax]

Too soon. The automatically triggered builds seem to start at about (our European) 8 am and take about 1-3 hours. Both are there now.

[kallisti5]

:-)