Comment 17 for bug 1655608

Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix merged to swift (master)

Reviewed: https://review.opendev.org/c/openstack/swift/+/788833
Committed: https://opendev.org/openstack/swift/commit/46ea3aeae817132957fb4fee737a1ad8d1e07837
Submitter: "Zuul (22348)"
Branch: master

commit 46ea3aeae817132957fb4fee737a1ad8d1e07837
Author: Alistair Coles <email address hidden>
Date: Thu Apr 29 20:48:51 2021 +0100

    Quarantine stale EC fragments after checking handoffs

    If the reconstructor finds a fragment that appears to be stale then it
    will now quarantine the fragment. Fragments are considered stale if
    insufficient fragments at the same timestamp can be found to rebuild
    missing fragments, and the number found is less than or equal to a new
    reconstructor 'quarantine_threshold' config option.

    Before quarantining a fragment the reconstructor will attempt to fetch
    fragments from handoff nodes in addition to the usual primary nodes.
    The handoff requests are limited by a new 'request_node_count'
    config option.

    'quarantine_threshold' defaults to zero i.e. no fragments will be
    quarantined. 'request node count' defaults to '2 * replicas'.

    Closes-Bug: 1655608

    Change-Id: I08e1200291833dea3deba32cdb364baa99dc2816