|
2 | 2 |
|
3 | 3 | All notable changes to this project are documented in this file.
|
4 | 4 |
|
| 5 | +## 0.23.0 |
| 6 | + |
| 7 | +**Release date:** 2022-04-12 |
| 8 | + |
| 9 | +This prerelease introduces new retention options for Garbage Collection, |
| 10 | +a new opt-in in-memory cache for `HelmRepository` index files, improves |
| 11 | +notifications following reconciling failures, brings ways to configure |
| 12 | +Key Exchange Algorithms, plus some extra housekeeping awesomeness. |
| 13 | + |
| 14 | +Garbage Collection is enabled by default, and now its retention options |
| 15 | +are configurable with the flags: `--artifact-retention-ttl` (default: `60s`) |
| 16 | +and `--artifact-retention-records` (default: `2`). They define the minimum |
| 17 | +time to live and the maximum amount of artifacts to survive a collection. |
| 18 | + |
| 19 | +A new notification is now emitted to identify recovery from failures. It |
| 20 | +is triggered when a failed reconciliation is followed by a successful one, and |
| 21 | +the notification message is the same that's sent in usual successful source |
| 22 | +reconciliation message about the stored artifact. |
| 23 | + |
| 24 | +The opt-in in-memory cache for `HelmRepository` addresses issues where the |
| 25 | +index file is loaded and unmarshalled in concurrent reconciliation resulting |
| 26 | +in a heavy memory footprint. It can be configured using the flags: |
| 27 | +`--helm-cache-max-size`, `--helm-cache-ttl`, `--helm-cache-purge-interval`. |
| 28 | + |
| 29 | +The Key Exchange Algorithms used when establishing SSH connections are |
| 30 | +based on the defaults configured upstream in `go-git` and `golang.org/x/crypto`. |
| 31 | +Now this can be overriden with the flag `--ssh-kex-algos`. Note this applies |
| 32 | +to the `go-git` gitImplementation or the `libgit2` gitImplementation but |
| 33 | +_only_ when Managed Transport is being used. |
| 34 | + |
| 35 | +Managed Transport for `libgit2` now introduces self-healing capabilities, |
| 36 | +to recover from failure when long-running connections become stale. |
| 37 | + |
| 38 | +The exponental back-off retry can be configured with the new flags: |
| 39 | +`--min-retry-delay` (default: `750ms`) and `--max-retry-delay` |
| 40 | +(default: `15min`). Previously the defaults were set to `5ms` and `1000s`, |
| 41 | +which in some cases impaired the controller's ability to self-heal |
| 42 | +(e.g. retrying failing SSH connections). |
| 43 | + |
| 44 | + |
| 45 | +Introduction of a secure directory loader which improves the handling |
| 46 | +of Helm charts paths. |
| 47 | + |
| 48 | +Improvements: |
| 49 | +- update toolkit.fluxcd.io docs links |
| 50 | + [#651](https://github.com/fluxcd/source-controller/pull/651) |
| 51 | +- Add optional in-memory cache of HelmRepository index files |
| 52 | + [#626](https://github.com/fluxcd/source-controller/pull/626) |
| 53 | +- Add flag to allow configuration of SSH kex algos |
| 54 | + [#655](https://github.com/fluxcd/source-controller/pull/655) |
| 55 | +- Garbage collect with provided retention options |
| 56 | + [#638](https://github.com/fluxcd/source-controller/pull/638) |
| 57 | +- Avoid event logging GC failure |
| 58 | + [#659](https://github.com/fluxcd/source-controller/pull/659) |
| 59 | +- Add notify() in all the reconcilers |
| 60 | + [#624](https://github.com/fluxcd/source-controller/pull/624) |
| 61 | +- Remove leftover timeout in reconcilers |
| 62 | + [#660](https://github.com/fluxcd/source-controller/pull/660) |
| 63 | +- libgit2: managed transport improvements |
| 64 | + [#658](https://github.com/fluxcd/source-controller/pull/658) |
| 65 | +- helm: introduce customized chart loaders |
| 66 | + [#663](https://github.com/fluxcd/source-controller/pull/663) |
| 67 | +- Add flags to configure exponential back-off retry |
| 68 | + [#664](https://github.com/fluxcd/source-controller/pull/664) |
| 69 | + |
5 | 70 | ## 0.22.5
|
6 | 71 |
|
7 | 72 | **Release date:** 2022-03-30
|
|
0 commit comments