|
| 1 | +--- |
| 2 | +layout: post |
| 3 | +title: "2019-10-29 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 Mark-Simulacrum. Minutes written by pietroalbini. |
| 9 | +Attending: alexcrichton, kennytm, Mark-Simulacrum, pietroalbini, shepmaster |
| 10 | +[Start of the conversation][discord] |
| 11 | + |
| 12 | +[discord]: https://discordapp.com/channels/442252698964721669/443148319431065610/638784152014946321 |
| 13 | + |
| 14 | +## News from the team |
| 15 | + |
| 16 | +- A big PR moving most of the CI build environment preparation to standalone |
| 17 | + scripts finally landed! This will allow multiple other CI improvements down |
| 18 | + the line! |
| 19 | + |
| 20 | +- A crates.io staging environment behind CloudFront has been configured on |
| 21 | + [staging.cratesio.com](https://staging.cratesio.com), testing the |
| 22 | + configuration before deploying it on the main domain. Please note the staging |
| 23 | + server is on a temporary domain and will be moved in the near future. |
| 24 | + |
| 25 | +## Rolling back the msys2 hack (P-high issue) |
| 26 | + |
| 27 | +Last week there was an issue in our CI caused by a broken `ca-certificates` |
| 28 | +msys2 package, that caused all the windows builders to fail. To patch the |
| 29 | +problem an older version of that package was vendored and CI was configured to |
| 30 | +download it instead of the broken one. Now that upstream fixed the issue, |
| 31 | +pietroalbini will prepare PRs targeting both master and beta rolling back the |
| 32 | +temporary patch. |
| 33 | + |
| 34 | +## What to do with mirroring msys2/mingw? |
| 35 | + |
| 36 | +The CI issue mentioned in the previous topic started thoughts about vendoring |
| 37 | +msys2 and MinGW as a whole, to prevent such issues from happening in the |
| 38 | +future. msys2 is a weird program to package as it’s more of a Linux-like |
| 39 | +distribution on Windows, with its own package manager based on tooling borrowed |
| 40 | +from Arch Linux. |
| 41 | + |
| 42 | +Creating the initial mirrors is not trivial but doable, but the concern is how |
| 43 | +to keep them updated, because for example an outdated OpenSSL is really bad. |
| 44 | +mati865 suggested that other projects often tar the whole msys2 installation |
| 45 | +directory and periodically update that, and we’ll explore that. More discussion |
| 46 | +on the approach will likely happen on the PR implementing the changes. |
| 47 | + |
| 48 | +## Restricted AWS access for docs.rs operators |
| 49 | + |
| 50 | +One of the issues exposed in the [docs.rs outage postmortem][outage] was that |
| 51 | +the people in the docs.rs on-call rotation did not have access to the AWS |
| 52 | +resources part of docs.rs. pietroalbini proposed to give them limited access, |
| 53 | +namely to stop/start the EC2 instance and to inspect/change the underlying S3 |
| 54 | +storage bucket. Mark-Simulacrum thought the access was not needed in practice, |
| 55 | +and we agreed to revisit the topic if issues arise again. |
| 56 | + |
| 57 | +[outage]: https://blog.rust-lang.org/inside-rust/2019/10/24/docsrs-outage-postmortem.html |
| 58 | + |
| 59 | +## New server for perf |
| 60 | + |
| 61 | +To ensure consistent results the collector for [perf.rust-lang.org] needs to be |
| 62 | +on bare metal hardware, and until now it was a server lying under |
| 63 | +alexcrichton’s desk. The server will have to be moved by December though, and |
| 64 | +alexcrichton asked the team on feedback on what to do. There was consensus that |
| 65 | +the best solution is to look for a rented bare metal server, and pietroalbini |
| 66 | +will do an initial investigation. |
| 67 | + |
| 68 | +[perf.rust-lang.org]: https://perf.rust-lang.org |
0 commit comments