ethereum
downloads-page not up to date
https://geth.ethereum.org/downloads showing latest as v1.14.5 (all platforms), so hasn't been updated in three releases - we're now at v1.14.8
For develop-builds, the last one is 1.14.6, https://github.com/ethereum/go-ethereum/tree/7cfff30b, somewhere around June 28, 2024 at 06:54 PM was the last successfull publish.
https://github.com/ethereum/go-ethereum/commits/master?since=2024-06-28&until=2024-07-03 -- the breakage happens suspiciously close to https://github.com/ethereum/go-ethereum/pull/30037
Unrelated tangent, we need to fix this at some point (from travis log, where we build a docker container):
>>> docker build --build-arg COMMIT=bd57f35f8d1e920ef2b234ea6cd1aa18715b358f --build-arg VERSION=1.14.9-unstable --build-arg BUILDNUM=30159 --tag ethereum/client-go:alltools-TAG -f Dockerfile.alltools .
DEPRECATED: The legacy builder is deprecated and will be removed in a future release.
Install the buildx component to build images with BuildKit:
https://docs.docker.com/go/buildx/
Interesting, from https://app.travis-ci.com/github/ethereum/go-ethereum/jobs/625078325 :
78dc4cc8d04a: Pushed
alltools-latest-arm64: digest: sha256:f8d72417d9a75653858be93cfb4bb77c4f70268f46824909369abd589c1a7ffd size: 951
ci.go:831: Build number mismatch on ethereum/client-go:latest-amd64: want 30159, have 30157
ci.go:831: Build number mismatch on ethereum/client-go:latest-amd64: want 30159, have 30157
ci.go:857: Relinquishing publish to other builder
The command "go run build/ci.go docker -image -manifest amd64,arm64 -upload ethereum/client-go" exited with 0.
This error just goes in the log, but the job is still "green"
EDIT: I think this is fine, it's the docker builder, and the docker builder has no problem
The building and updload to azure appears to work fine: https://app.travis-ci.com/github/ethereum/go-ethereum/jobs/625078326#L5684 .
So maybe the upload is not the problem, but rather the website regeneration?
Yes, the blobs are there
$ curl "https://gethstore.blob.core.windows.net/builds?restype=container&comp=list&maxresults=10&prefix=geth-linux-amd64-1.14.8"
<?xml version="1.0" encoding="utf-8"?><EnumerationResults ServiceEndpoint="https://gethstore.blob.core.windows.net/" ContainerName="builds"><Prefix>geth-linux-amd64-1.14.8</Prefix><MaxResults>10</MaxResults><Blobs><Blob><Name>geth-linux-amd64-1.14.8-a9523b64.tar.gz</Name><Properties><Last-Modified>Mon, 12 Aug 2024 13:00:37 GMT</Last-Modified><Etag>0x8DCBACEC30E9F8F</Etag><Content-Length>29994640</Content-Length><Content-Type>application/octet-stream</Content-Type><Content-Encoding /><Content-Language /><Content-MD5>YRV3bM4ymlwo1+/82qxe/Q==</Content-MD5><Cache-Control /><Content-Disposition /><BlobType>BlockBlob</BlobType><LeaseStatus>unlocked</LeaseStatus><LeaseState>available</LeaseState></Properties></Blob><Blob><Name>geth-linux-amd64-1.14.8-a9523b64.tar.gz.asc...
This seems to be the issue: https://app.netlify.com/sites/geth-website/deploys/66ad0e7d8182b00008e98725#L113
Broken by https://github.com/ethereum/go-ethereum/pull/30101
I don't mind that we sometime break the build, but I find it a bit scary that we had broken production deployments for six weeks but didn't get any notification. Perhaps we can provision netlify with some email address to report to?
Thanks for raising @holiman. I've added an email notification to the website team in Netlify for any failed deployments. Limited bandwidth though for ongoing management, so let me know if you'd like any alerts directed your way as well (@ me please), thanks!
Agreed! Perhaps also don't allow PRs to be merged if the Netlify build fails?
General FYI - I know no longer work on the ethereum.org team so I'm not be available to help with the website. Much love ✌️😎
Perhaps also don't allow PRs to be merged if the Netlify build fails?
IIRC, that won't work. All PRs from external branches do fail the netlify build -- only builds from branches on this repo succeeds. So we don't really want to be notified of every fail, what's important is if the deployment of the merged website branch fails. That is, the prod deployment.
General FYI - I know no longer work on the ethereum.org team so I'm not be available to help with the website
Right, sorry for the noise! Cheers!
Sorry, going to have to reopen this.
https://geth.ethereum.org/downloads -- the develop builds are showing the latest being Aug 15, 6 days ago. There are actually later binaries, so it appears the site hasn't been cron-updating.
The deploy-page https://app.netlify.com/sites/geth-website/deploys shows latest production deploy Aug 16, nothing since.
I had supposed it should run nightly? Is that disabled?
I had supposed it should run nightly? Is that disabled?
Ah, no it's not regenerated in full, rather some "incremental static regeneration" of the data on the downloads-page:
https://github.com/ethereum/go-ethereum/blob/website/src/pages/downloads.tsx#L37
The latest deploy log does mention the ISR: https://app.netlify.com/sites/geth-website/deploys/66befd84c18f230008a49946#L134
I'm not sure what to make of this. "ISR is great and awesome. OBS: Not supported though" ... what? Are we using edge? Strange
Interesting, from https://app.travis-ci.com/github/ethereum/go-ethereum/jobs/625078325 :
78dc4cc8d04a: Pushed alltools-latest-arm64: digest: sha256:f8d72417d9a75653858be93cfb4bb77c4f70268f46824909369abd589c1a7ffd size: 951 ci.go:831: Build number mismatch on ethereum/client-go:latest-amd64: want 30159, have 30157 ci.go:831: Build number mismatch on ethereum/client-go:latest-amd64: want 30159, have 30157 ci.go:857: Relinquishing publish to other builder The command "go run build/ci.go docker -image -manifest amd64,arm64 -upload ethereum/client-go" exited with 0.This error just goes in the log, but the job is still "green"
EDIT: I think this is fine, it's the docker builder, and the docker builder has no problem
Just to answer this bit of the ticket. To build a multi-architecture docker image, we need to build each on it's own platform, and then have a 3rd job that merges the two images into a multi-arch image.
To avoid having 3 travis jobs and a weird dependency across them, I did a dirty trick in the amd/arm64 docker builders: after they push their own architecture image, they check if the couterpart is already pushed or not. If not (i.e. mismatching build number upstream), then we early exit as the sibling builder is still running. If the counterpart image matches our build numbers, then our sibling already quit and it's up to us to combine the two images into the multi-arch one.
The error message I think is simple a log saying that "we finished first, exit and leave the other builder to do the multi-arch assembly".
