Ceph Monitor handling

Working in a customer project with RHEL OSP 6 and CEPH Enterprise is somehow challenging Here are some experiences especially with CEPH Monitor.

 

Ceph monitor start / stop

Ceph monitor on RHEL 7 adds a SysV start script in /etc/init.d/ named “ceph”. You can verify it with “chkconfig –list”. There you should find an entry for “ceph”. Ensure, that the service is enabled.

To start the Ceph monitor manually perform the following:


service ceph start

To find out the do the following. This will print out the name of the monitor and its status:

service ceph status

Ceph monitor failed to start

Due to a lot of reasons a Ceph monitor start could fail (for example clock skew > 0.05 seconds etc). In our case the one of our monitors did not start because he did not receive the correct monitor map. Bringing back an already joined monitor to the cluster perform the following:

1) Check ceph health to find out which monitor failed

ceph health detail

2) On the failed monitor perform the following:

Print out the current cluster monitor map

ceph mon dump

Then get the current cluster monitor map and save it:

ceph mon getmap -o /tmp/monmap

Verify that all configured monitors are present in the map:

monmaptool --print /tmp/monmap

Then inject the new monitor map on your failed monitor server:

ceph-mon -i --inject-monmap /tmp/monmap

In my case -i was the shortname of the monitor hostname.

Now it`s time to restart the monitor service (take a look at chapter “Ceph monitor start/stop”)

service ceph start

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.