Ceph is high performance distributed storage system, for storing petabytes of data on commodity hardware. Agent uses Ceph built in tools to expose statistics and send to OddEye servers. Configuration is very minimal and easy : All that is needed is to tell Agent the username, which is allowed to grab statistics (Typically client.admin) and path to keyring. If you run Ceph, you already have installed all needed dependencies, so nothing else is needed to add. Agent will look for configuration in file bigdata.ini.

cd ${OE_AGENT_HOME}/checks_enabled
ln -s ../checks_available/check_ceph.py ./

Make sure that /etc/ceph/ceph.client.admin.keyring is readable for user, which runs Agent. If you want to use another keyring or ceph user, change client and keyring paramaters to your desired ones. It takes cople of seconds, afterwords you can create graphs with these metrics :

client: client.admin
keyring: /etc/ceph/ceph.client.admin.keyring
${OE_AGENT_HOME}/oddeye.sh restart
Name Description Type Unit
ceph_num_bytes Amount of free space in Ceph gauge Bytes
ceph_num_pgs Total amount of placement groups gauge None
ceph_raw_bytes Total raw space for ceph cluster gauge Bytes
ceph_raw_bytes_avail Amount of used raw available replication is taken to account gauge Bytes
ceph_raw_bytes_used Amount of used raw space replication is taken to account gauge Bytes
ceph_io_sec I/O operations per second for entire cluster rate OPS
ceph_read_bytes_sec Read bytes per second for entire cluster rate Bytes
ceph_write_bytes_sec Write bytes per second for entire cluster rate Bytes
ceph_degraded_objects Number of degraded objects gauge None
ceph_degraded_ratio Ratio of degraded objects gauge Float
ceph_degraded_total Number of total degraded objects gauge None
ceph_recovering_bytes_per_sec Speed of recovery in bytes gauge Bytes
ceph_num_objects_recovered Recovered objects gauge None
ceph_recovering_keys_per_sec Recovered Keys rate None
ceph_degraded_percent Percentage of degraded objects gauge Percent
ceph_misplaced_percent Percentage of misplaced objects gauge Percent

There isn no necessity to run check_ceph.py on all nodes of Ceph of cluster. Statistics from all nodes will be the same, as above mentioned stats are global at cluster level.