Skip to content

v0.65.0

Choose a tag to compare

@b10x-bot b10x-bot released this 02 Nov 15:27
· 31 commits to main since this release
8150245

connectors 0.65.0

Fixed:

  • connections connect --endpoint no longer blames the endpoint when the catalog provider
    installed beside the CLI cannot run. A provider that is missing, not executable, killed,
    exited, past its deadline or printing nothing usable answers unavailable at stage
    installed_provider, with installed_provider_path and installed_provider_failure; setup check
    lists the installed_provider prerequisite. An endpoint a running provider refuses stays
    invalid_input on endpoint.
  • setup init writes connectors-local/5, and setup migrate --apply moves a connectors-local/2
    file with no legacy entries to /5, with a backup.

Added:

  • instances status names why a program failed to start: start_failure, start_exit_code or
    start_signal, and start_log, an owner-only file holding the first 64 KiB the program wrote
    to standard error before its handshake ended.
  • The release archives carry connectors-grafana, connectors-prometheus, connectors-loki and
    connectors-kubernetes for every target, Apple-silicon macOS included; each answers --version,
    and a macOS CI job lints and tests them.
  • A Kubernetes exec credential plugin, such as aws eks get-token, runs on macOS from its pinned
    path after the file, digest and directory-ownership checks.
    Full list: https://github.com/beyond10x/connectors/blob/v0.65.0/CHANGELOG.md

Limitations:

  • The macOS binaries are not signed or notarized; after download run
    xattr -dr com.apple.quarantine on the unpacked directory.
  • On macOS, root or the same user can replace an exec plugin between its check and its run;
    Linux runs the file it hashed.
  • A Grafana instance that never started on an adopter's Mac was not reproduced in CI, where the
    Grafana program starts and its tests pass; start_failure and start_log now name the cause.