|
| 1 | +--- |
| 2 | +layout: post |
| 3 | +title: "2019-11-12 Infrastructure Team Meeting" |
| 4 | +author: Pietro Albini |
| 5 | +team: the infrastructure team <https://www.rust-lang.org/governance/teams/operations#infra> |
| 6 | +--- |
| 7 | + |
| 8 | +Meeting run by pietroalbini. Minutes written by pietroalbini. |
| 9 | +Attending: alexcrichton, kennytm, Mark-Simulacrum, pietroalbini, sgrif, shepmaster |
| 10 | +[Start of the conversation](https://discordapp.com/channels/442252698964721669/443148319431065610/643872655312683018) |
| 11 | + |
| 12 | +## Rust 1.14.0 Cargo artifacts on the rust-lang-ci S3 bucket |
| 13 | + |
| 14 | +Following on last meeting’s topics, pietroalbini investigate whether it would |
| 15 | +be possible to redirect requests for Cargo 1.14.0 from the `rust-lang-ci` S3 |
| 16 | +bucket to our CDN. The answer is no, as the S3 support for redirects requires a |
| 17 | +different endpoint than the one included in the manifest. The two main courses |
| 18 | +of action were to nothing (keeping the files in the `rust-lang-ci` S3 bucket), |
| 19 | +or to rewrite and resign the manifest of Rust 1.14.0 to point to the CDN. |
| 20 | + |
| 21 | +Mark-Simulacrum was in favor of doing nothing, but pietroalbini pointed out |
| 22 | +keeping releases in two different places will make mirroring and caching |
| 23 | +efforts more complex, as mirror authors will have to special-case a separate |
| 24 | +domain just for a release. We then decided to rewrite and resing the manifest. |
| 25 | + |
| 26 | +shepmaster proposed to add a test somewhere to ensure this doesn’t happen |
| 27 | +again, but nobody on the team had the time to do this. If someone is interested |
| 28 | +please ask in the infra channel. |
| 29 | + |
| 30 | +## Deduplicating CI configuration on GitHub Actions |
| 31 | + |
| 32 | +One issue about GitHub Actions is it doesn’t allow to include shared pieces of |
| 33 | +configuration into workflows, forcing us to duplicate (for example) the steps |
| 34 | +needed to complete a build. pietroalbini investigated a few ways to work around |
| 35 | +the limitation, and the options he presented during the meeting are either |
| 36 | +writing our own configuration format and having a tool generate the GitHub |
| 37 | +Actions configuration from it, or duplicating the configuration manually and |
| 38 | +having a tool to ensure the manually duplicated things don’t diverge. |
| 39 | + |
| 40 | +The rest of the team expressed concerns with generating the configuration, as |
| 41 | +using the generator adds even more complexity to our already complex CI. On the |
| 42 | +other hand pietroalbini didn’t like the tool to ensure the configuration |
| 43 | +doesn’t diverge, as that would make changes to the CI configuration harder for |
| 44 | +the people doing it. The agreement in the meeting is that pietroalbini will try |
| 45 | +to create other proof of concepts, hoping to find a better solution. |
| 46 | + |
| 47 | +## New server for perf |
| 48 | + |
| 49 | +alexcrichton didn’t hear back from Hetzner about the new perf server yet. |
0 commit comments