openstack/tripleo-docs
Code Issues Proposed changes

Document how to quiesce and delete a cephstorage node

Browse Source 
Adds a chapter in post_deployment describing how to cleanly remove
from the Ceph cluster a cephstorage node and all the OSDs it hosted.

Change-Id: I6690bb54e4724d1042ecc3fa812cc390de6ce49d
This commit is contained in:
Giulio Fidente 2016-11-29 22:49:16 +01:00
parent d927b28a10
commit 39b88b1ff4
3 changed files with 60 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
doc/source/post_deployment/delete_nodes.rst
View File

@ -18,8 +18,9 @@ IDs (which represent nodes) to be deleted.
changes to the overcloud.
.. note::
Before deleting a compute node please make sure that the node is quiesced,
see :ref:`quiesce_compute`.
Before deleting a compute node or a cephstorage node, please make sure that
the node is quiesced, see :ref:`quiesce_compute` or
:ref:`quiesce_cephstorage`.
.. note::
A list of nova instance IDs can be listed with command::

1
doc/source/post_deployment/post_deployment.rst
View File

@ -10,6 +10,7 @@ In this chapter you will find advanced management of various |project| areas.
scale_roles
delete_nodes
quiesce_compute
quiesce_cephstorage
vm_snapshot
package_update
upgrade

56
doc/source/post_deployment/quiesce_cephstorage.rst Normal file
View File

@ -0,0 +1,56 @@
.. _quiesce_cephstorage:
Quiescing a CephStorage Node
============================
The process of quiescing a cephstorage node means to inform the Ceph
cluster that one or multiple OSDs will be permanently removed so that
the node can be shut down without affecting the data availability.
Take the OSDs out of the cluster
--------------------------------
Before you remove an OSD, you need to take it out of the cluster so that Ceph
can begin rebalancing and copying its data to other OSDs. Running the following
commands on a given cephstorage node will take all data out of the OSDs hosted
on it::
OSD_IDS=$(ls /var/lib/ceph/osd | awk 'BEGIN { FS = "-" } ; { print $2 }')
for OSD_ID in $OSD_IDS; do ceph crush reweight osd.$OSD_ID 0.0; done
Ceph will begin rebalancing the cluster by migrating placement groups out of
the OSDs. You can observe this process with the ceph tool::
ceph -w
You should see the placement group states change from active+clean to active,
some degraded objects, and finally active+clean when migration completes.
Removing the OSDs
-----------------
After the rebalancing, the OSDs will still be running. Running the following on
that same cephstorage node will stop all OSDs hosted on it, remove them from the
CRUSH map, from the OSDs map and delete the authentication keys::
OSD_IDS=$(ls /var/lib/ceph/osd | awk 'BEGIN { FS = "-" } ; { print $2 }')
for OSD_ID in $OSD_IDS; do
ceph osd out $OSD_ID
systemctl stop ceph-osd@$OSD_ID
ceph osd crush remove osd.$OSD_ID
ceph auth del osd.$OSD_ID
ceph osd rm $OSD_ID
done
.. admonition:: Mitaka
:class: mitaka
TripleO/Mitaka uses and supports Ceph/Hammer, not Jewel, which does not
use systemd but sysv init scripts. For Mitaka the systemctl command above
which stops the OSD should be replaced by::
service ceph stop osd.$OSD_ID
You are now free to reboot or shut down the node (using the Ironic API), or
even remove it from the overcloud altogether by scaling down the overcloud
deployment, see :ref:`delete_nodes`.