Some Ceph cluster stats not reported for Hammer

[moorglade]

Relevant telegraf.conf:

[[inputs.ceph]]
  ceph_user = "client.admin"
  ceph_config = "/etc/ceph/ceph.conf"
  gather_admin_socket_stats = false
  gather_cluster_stats = true

System info:

Telegraf 1.9.5 (executed in a Docker container)
Ceph 0.94.5 (Hammer)
CentOS 7.5.1804

Steps to reproduce:

1. Add Ceph input with gather_cluster_stats = true to Telegraf configuration.

Expected behavior:

All cluster stats described in the README are reported.

Actual behavior:

Some cluster stats are not reported at all, i.e.:

• ceph_usage.total_bytes
• ceph_usage.total_used_bytes
• ceph_usage.total_avail_bytes
• ceph_pool_usage.max_avail

(There may be more, I haven't checked all the metrics.)

Additional info:

1. Everything works fine with Telegraf 1.8.3.

2. This issue seem to be related to ceph input plugin cluster metrics broken on Luminous #5277, but instead of zeroes, I don't get any values at all.

3. Output returned by ceph df seems to be partially consistent with what I can see in the tests, i.e.

   $ ceph --conf /etc/ceph/ceph.conf --format json-pretty df
   {
       "stats": {
           "total_bytes": 56719207284736,
           "total_used_bytes": 26185807761408,
           "total_avail_bytes": 30533399523328
       },
       "pools": [
           {
               "name": "rbd",
               "id": 0,
               "stats": {
                   "kb_used": 1172409821,
                   "bytes_used": 1200547656115,
                   "max_avail": 13376747995136,
                   "objects": 294000
               }
           },
   ...
   

[moorglade]

May I ask which release is this fix planned for? I see it's been marked for milestone 1.10.0, but I still don't see it in the 1.10.1 release.

[danielnelson]

It looks like I forgot to include it into the release branch, I will put it in for 1.10.2.

[moorglade]

Thanks!