Chapter 10. Troubleshooting clusters in stretch mode

Procedure

1. Disable automated monitor deployment:

   Example

   [ceph: root@host01 /]# ceph orch apply mon --unmanaged
   
   Scheduled mon update…

2. View the monitors in quorum:

   Example

   [ceph: root@host01 /]# ceph -s
   
   mon: 5 daemons, quorum host01, host02, host04, host05 (age 30s), out of quorum: host07

3. Set the monitor in quorum as a new tiebreaker:

   Syntax

   ceph mon set_new_tiebreaker NEW_HOST

   Example

   [ceph: root@host01 /]# ceph mon set_new_tiebreaker host02

   Important

   You get an error message if the monitor is in the same location as existing non-tiebreaker monitors:

   Example

   [ceph: root@host01 /]# ceph mon set_new_tiebreaker host02
   
   Error EINVAL: mon.host02 has location DC1, which matches mons host02 on the datacenter dividing bucket for stretch mode.

   If that happens, change the location of the monitor:

   Syntax

   ceph mon set_location HOST datacenter=DATACENTER

   Example

   [ceph: root@host01 /]# ceph mon set_location host02 datacenter=DC3

4. Remove the failed tiebreaker monitor:

   Syntax

   ceph orch daemon rm FAILED_TIEBREAKER_MONITOR --force

   Example

   [ceph: root@host01 /]# ceph orch daemon rm mon.host07 --force
   
   Removed mon.host07 from host 'host07'

5. Once the monitor is removed from the host, redeploy the monitor:

   Syntax

   ceph mon add HOST IP_ADDRESS datacenter=DATACENTER
   ceph orch daemon add mon HOST

   Example

   [ceph: root@host01 /]# ceph mon add host07 213.222.226.50 datacenter=DC1
   [ceph: root@host01 /]# ceph orch daemon add mon host07

6. Ensure there are five monitors in quorum:

   Example

   [ceph: root@host01 /]# ceph -s
   
   mon: 5 daemons, quorum host01, host02, host04, host05, host07 (age 15s)

7. Verify that everything is configured properly:

   Example

   [ceph: root@host01 /]# ceph mon dump
   
   epoch 19
   fsid 1234ab78-1234-11ed-b1b1-de456ef0a89d
   last_changed 2023-01-17T04:12:05.709475+0000
   created 2023-01-16T05:47:25.631684+0000
   min_mon_release 16 (pacific)
   election_strategy: 3
   stretch_mode_enabled 1
   tiebreaker_mon host02
   disallowed_leaders host02
   0: [v2:132.224.169.63:3300/0,v1:132.224.169.63:6789/0] mon.host02; crush_location {datacenter=DC3}
   1: [v2:220.141.179.34:3300/0,v1:220.141.179.34:6789/0] mon.host04; crush_location {datacenter=DC2}
   2: [v2:40.90.220.224:3300/0,v1:40.90.220.224:6789/0] mon.host01; crush_location {datacenter=DC1}
   3: [v2:60.140.141.144:3300/0,v1:60.140.141.144:6789/0] mon.host07; crush_location {datacenter=DC1}
   4: [v2:186.184.61.92:3300/0,v1:186.184.61.92:6789/0] mon.host03; crush_location {datacenter=DC2}
   dumped monmap epoch 19

8. Redeploy the monitors:

   Syntax

   ceph orch apply mon --placement="HOST_1, HOST_2, HOST_3, HOST_4, HOST_5”

   Example

   [ceph: root@host01 /]# ceph orch apply mon --placement="host01, host02, host04, host05, host07"
   
   Scheduled mon update...

Procedure

1. Add a new monitor to the cluster:

   1. Manually add the crush_location to the new monitor:

      Syntax

      ceph mon add NEW_HOST IP_ADDRESS datacenter=DATACENTER

      Example

      [ceph: root@host01 /]# ceph mon add host06 213.222.226.50 datacenter=DC3
      
      adding mon.host06 at [v2:213.222.226.50:3300/0,v1:213.222.226.50:6789/0]

      Note

      The new monitor has to be in a different location than existing non-tiebreaker monitors.

   2. Disable automated monitor deployment:

      Example

      [ceph: root@host01 /]# ceph orch apply mon --unmanaged
      
      Scheduled mon update…

   3. Deploy the new monitor:

      Syntax

      ceph orch daemon add mon NEW_HOST

      Example

      [ceph: root@host01 /]# ceph orch daemon add mon host06

2. Ensure there are 6 monitors, from which 5 are in quorum:

   Example

   [ceph: root@host01 /]# ceph -s
   
   mon: 6 daemons, quorum host01, host02, host04, host05, host06 (age 30s), out of quorum: host07

3. Set the new monitor as a new tiebreaker:

   Syntax

   ceph mon set_new_tiebreaker NEW_HOST

   Example

   [ceph: root@host01 /]# ceph mon set_new_tiebreaker host06

4. Remove the failed tiebreaker monitor:

   Syntax

   ceph orch daemon rm FAILED_TIEBREAKER_MONITOR --force

   Example

   [ceph: root@host01 /]# ceph orch daemon rm mon.host07 --force
   
   Removed mon.host07 from host 'host07'

5. Verify that everything is configured properly:

   Example

   [ceph: root@host01 /]# ceph mon dump
   
   epoch 19
   fsid 1234ab78-1234-11ed-b1b1-de456ef0a89d
   last_changed 2023-01-17T04:12:05.709475+0000
   created 2023-01-16T05:47:25.631684+0000
   min_mon_release 16 (pacific)
   election_strategy: 3
   stretch_mode_enabled 1
   tiebreaker_mon host06
   disallowed_leaders host06
   0: [v2:213.222.226.50:3300/0,v1:213.222.226.50:6789/0] mon.host06; crush_location {datacenter=DC3}
   1: [v2:220.141.179.34:3300/0,v1:220.141.179.34:6789/0] mon.host04; crush_location {datacenter=DC2}
   2: [v2:40.90.220.224:3300/0,v1:40.90.220.224:6789/0] mon.host01; crush_location {datacenter=DC1}
   3: [v2:60.140.141.144:3300/0,v1:60.140.141.144:6789/0] mon.host02; crush_location {datacenter=DC1}
   4: [v2:186.184.61.92:3300/0,v1:186.184.61.92:6789/0] mon.host05; crush_location {datacenter=DC2}
   dumped monmap epoch 19

6. Redeploy the monitors:

   Syntax

   ceph orch apply mon --placement="HOST_1, HOST_2, HOST_3, HOST_4, HOST_5”

   Example

   [ceph: root@host01 /]# ceph orch apply mon --placement="host01, host02, host04, host05, host06"
   
   Scheduled mon update…

Procedure

1. Force the stretch cluster into the recovery mode:

   Example

   [ceph: root@host01 /]#  ceph osd force_recovery_stretch_mode --yes-i-really-mean-it

   Note

   The recovery state puts the cluster in the HEALTH_WARN state.

2. When in recovery mode, the cluster should go back into normal stretch mode after the placement groups are healthy. If that does not happen, you can force the stretch cluster into the healthy mode:

   Example

   [ceph: root@host01 /]#  ceph osd force_healthy_stretch_mode --yes-i-really-mean-it

   Note

   You can also run this command if you want to force the cross-data-center peering early and you are willing to risk data downtime, or you have verified separately that all the placement groups can peer, even if they are not fully recovered.

   You might also wish to invoke the healthy mode to remove the HEALTH_WARN state, which is generated by the recovery state.

   Note

   The force_recovery_stretch_mode and force_recovery_healthy_mode commands should not be necessary, as they are included in the process of managing unanticipated situations.