Removing Purged OSDs in Ceph

News Knowledge Base
1

Share the steps to remove purged OSDs.

This article provides step-by-step instructions to safely remove a purged OSD (Object Storage Daemon) from a Ceph cluster using rook-ceph-tools and Ceph CLI.

Prerequisites

  • kubectl CLI configured to interact with the cluster.
  • Ensure that the OSD to be removed is no longer required or has failed permanently.

Steps

Follow these steps to remove a purged OSD (osd.{number}):

  1. Mark the OSD as Out - This step ensures that the Placement Groups (PGs) are moved off the OSD. 
    ceph osd out osd.{number}

    1. Monitor PG Remapping - Verify that PGs have been successfully remapped and no data remains on the OSD. 
      kubectl -n rook-ceph exec deploy/rook-ceph-tools -- ceph osd df

      💡 Wait until osd.{number} shows no data and its PG count is 0.

      1. Purge the OSD 
        ceph osd purge {number} --yes-i-really-mean-it

        This command removes the OSD and its metadata from the cluster.

        1. Remove from CRUSH Map 
          ceph osd crush remove osd.{number}

          1. Delete OSD from OSD Map 
            ceph osd rm osd.{number}

            1. Remove OSD Authentication 
              ceph auth del osd.{number}

              Notes

              • Replace {number} with the actual OSD ID (e.g., 0, 1, 2, etc.).
              • Always ensure PGs are fully rebalanced before purging an OSD to avoid data loss.