rocksdb-resharding.yml

---
# Copyright Red Hat
# SPDX-License-Identifier: Apache-2.0
# Author: Guillaume Abrioux <gabrioux@redhat.com>
#
# This playbook reshards the rocksDB database for a given OSD
#
# Usage:
#
# ansible-playbook -i <inventory host file> rocksdb-resharding.yml -e osd_id=0 -e admin_node=ceph-mon0
#
# Required run-time variables
# ------------------
# osd_id : the id of the OSD where you want to reshard its corresponding rocksdb database.
# admin_node : the name of a node with enough privileges to stop/start
#              daemons via `cephadm shell ceph orch daemon` command.
#              (usually the bootstrap node).
#
# Optional run-time variables
# ------------------
# fsid : the fsid of the cluster.
# rocksdb_sharding_parameters : the rocksdb sharding parameter to set. Default is 'm(3) p(3,0-12) O(3,0-13) L P'.
# docker : bool to be set in order to use docker engine instead. Default is False.

- hosts: localhost
  become: true
  gather_facts: false
  tasks:
    - name: fail if osd_id is not defined
      fail:
        msg: "you must provide 'osd_id' variable"
      when: osd_id is undefined

    - name: fail if admin_node is not defined
      fail:
        msg: "you must pass 'admin_node' variable"
      when: admin_node is not defined

    - name: test connectivity to admin node
      ping:
      delegate_to: "{{ admin_node }}"

    - name: fail if osd_id isn't an id
      fail:
        msg: "osd_id must be an id"
      when: not osd_id is regex('^\d+$')

    - name: set_fact cephadm_cmd
      set_fact:
        cephadm_cmd: "cephadm {{ '--docker' if docker | default(False) | bool else '' }} shell ceph"

    - name: get details about the osd daemon
      delegate_to: "{{ admin_node }}"
      block:
        - name: get cluster fsid
          command: "{{ cephadm_cmd }} fsid"
          register: fsid
          changed_when: false
          when: fsid is not defined

        - name: set_fact fsid
          set_fact:
            fsid: "{{ fsid.stdout }}"
          when: fsid.stdout is defined

        - name: get container image currently used by osd container
          command: "{{ cephadm_cmd }} orch ps --daemon_type osd --daemon_id {{ osd_id }} --format json"
          changed_when: false
          register: ceph_orch_ps

        - name: set_fact container_image, container_host
          set_fact:
            container_image: "{{ (ceph_orch_ps.stdout | from_json)[0]['container_image_name'] }}"
            container_host: "{{ (ceph_orch_ps.stdout | from_json)[0]['hostname'] }}"

        - name: stop the osd
          ceph_orch_daemon:
            fsid: "{{ fsid }}"
            state: stopped
            daemon_id: "{{ osd_id }}"
            daemon_type: osd

        - name: set_fact ceph_cmd
          set_fact:
            ceph_bluestore_tool_cmd: "{{ container_binary | default('podman') }} run --rm --privileged --entrypoint=ceph-bluestore-tool -v /var/run/ceph/{{ fsid }}:/var/run/ceph:z -v /var/log/ceph/{{ fsid }}:/var/log/ceph:z -v /var/lib/ceph/{{ fsid }}/crash:/var/lib/ceph/crash:z -v /var/lib/ceph/{{ fsid }}/osd.{{ osd_id }}:/var/lib/ceph/osd/ceph-{{ osd_id }}:z -v /var/lib/ceph/{{ fsid }}/osd.{{ osd_id }}/config:/etc/ceph/ceph.conf:z -v /dev:/dev -v /run/udev:/run/udev -v /sys:/sys -v /var/lib/ceph/{{ fsid }}/selinux:/sys/fs/selinux:ro -v /run/lvm:/run/lvm -v /run/lock/lvm:/run/lock/lvm {{ container_image }} --path /var/lib/ceph/osd/ceph-{{ osd_id }}"

    - name: resharding operations
      delegate_to: "{{ container_host }}"
      block:
        - name: check fs consistency with fsck before resharding
          command: "{{ ceph_bluestore_tool_cmd }} fsck"
          changed_when: false

        - name: show current sharding
          command: "{{ ceph_bluestore_tool_cmd }} show-sharding"
          changed_when: false

        - name: reshard
          command: "{{ ceph_bluestore_tool_cmd }} --sharding=\"{{ rocksdb_sharding_parameters | default('m(3) p(3,0-12) O(3,0-13) L P') }}\" reshard"
          changed_when: false

        - name: check fs consistency with fsck after resharding
          command: "{{ ceph_bluestore_tool_cmd }} fsck"
          changed_when: false

    - name: restart the osd
      ceph_orch_daemon:
        fsid: "{{ fsid }}"
        state: started
        daemon_id: "{{ osd_id }}"
        daemon_type: osd
      delegate_to: "{{ admin_node }}"