Skip to content

SHIELD v8.2.0

Compare
Choose a tag to compare
@starkandwayne-bot starkandwayne-bot released this 22 May 16:37
· 295 commits to master since this release

Bug Fixes

  • The shieldd binary now properly reports its release version in
    both CLI (-v) and web UI contexts.

  • The archives list on the system page now no longer gives you the
    option of restoring invalid archives (i.e. purged stuff).
    Thanks @thomasmitchell for finding and reporting in #506.

  • System-initiated archive purges now properly set the store agent
    for purgation, so that the purge task has someone to talk to for
    removal of the archive from cloud storage. See #514.

  • The core scheduler now immediately fails any task for which the
    remote SHIELD agent does not signal a successful (rc=0) exit
    status. This should clean up some task logs, and remove red
    herring issues like JSON unmarshal failures, while
    simultaneously ensuring that failed purge tasks are re-tried.
    See #518.

  • Purge tasks are now being properly supplied with the restore key
    necessary for deleting the archive blob. See #516.

  • Agent Status tasks (op agent-status) were not previously being
    created with proper global tenant association. This prohibited
    operators from viewing the details of those tasks. We fixed
    this, and added a data fixup created to re-associate existing
    tasks. See #522.

  • The HUD now always registers the global cloud storage in its
    health data, so operators are aware of all issues with storaage
    systems that they might be using, global or tenant-private.
    See #504.

  • Jobs created via the Web UI now properly set their "KeepN"
    attribute, which was missing from the ingestion / insertion.
    Accompanying this is a new data fixup that should re-calculate
    the keep_n database field wherever it is zero. See #460.

  • Unlocking SHIELD was not properly giving feedback when a wrong
    password was given or no password at all. This is now working.
    See #524.

  • Previously we had issues doing a fs backup and restore of SHIELD
    itself. Improvments to the fs plugin, and fixing permissions issues
    within core now allow SHIELD to properly backup and restore itself.

Improvements

  • All -v handlers in CLI utilities now properly handle the 'dev'
    version as analogous to the empty ('') version, and revert to
    reporting the version of the binary as '(development)'. This is
    mainly for packaing Docker images properly.

  • All shield* CLI utilities, include the shield CLI itself,
    the shieldd daemon, and all helper binaries now sport options
    for getting their usage (--help) and versions (--versions).

  • The s3 plugin now accepts a URL as its s3_host endpoint
    parameter, affording operators more flexibility.
    The alternative was confusion! See #509.

  • When purging archives manually, you can now supply
    human-friendly reasons for the purge. For example, if the data
    is known to be bad in that particular vintage of the target
    system, you can purge the archives containing it, and explain
    that. See #520

  • Archives can now be annotated from the command-line, with the
    annotate-archive command.

  • Manually purged archives now track their reason for purge as
    "manually purged", instead of "expired". See #517.

  • All system- and tenant-level objects can now be searched for,
    and referenced by short UUIDs. This is huge (though short),
    going a long way to making the CLI easier to work with.

Release Engineering

  • Docker images can now be built with embedded release versions,
    for non-dev distribution as a container image.