Re: ceph fs perf stats output is empty

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Each array has different types. You need to look at the below cephfs-top code and see how they are interpreted.

[1] https://github.com/ceph/ceph/blob/main/src/tools/cephfs/top/cephfs-top#L66-L83 [2] https://github.com/ceph/ceph/blob/main/src/tools/cephfs/top/cephfs-top#L641-L714 [3] https://github.com/ceph/ceph/blob/main/src/tools/cephfs/top/cephfs-top#L110-L143 [4] https://github.com/ceph/ceph/blob/main/src/tools/cephfs/top/cephfs-top#L775-L823

On 12/06/23 14:23, Denis Polom wrote:

Hi,

yes, I've found that trick was I had to wait for about 15 sec to see the metrics.

Now I can see some numbers. Are units there in miliseconds? And also I see 2 numbers reported - is the first value actual and second is delta?

      "client.4636": [
        [
          924,
          4
        ],
        [
          0,
          0
        ],
        [
          10,
          981017512
        ],
        [
          0,
          33484910
        ],
        [
          136,
          0
        ],
        [
          1,
          2
        ],
        [
          2,
          2
        ],
        [
          1,
          2
        ],
        [
          0,
          0
        ],
        [
          302,
          1266679808
        ],
        [
          0,
          0
        ],
        [
          0,
          0
        ],
        [
          0,
          36361015
        ],
        [
          4205537661271535,
          302
        ],
        [
          0,
          11161636
        ],
        [
          190208004421472,
          3
        ]
      ]
    }
  },

Thx!



On 6/12/23 06:36, Jos Collin wrote:
Additionally, the first `ceph fs perf stats` output would be empty or outdated in most cases. You need to query a few times to get the latest values. So try `watch ceph fs perf stats`.

On Mon, 12 Jun 2023 at 06:30, Xiubo Li <xiubli@xxxxxxxxxx> wrote:


    On 6/10/23 05:35, Denis Polom wrote:
    > Hi
    >
    > I'm running latest Ceph Pacific 16.2.13 with Cephfs. I need to
    collect
    > performance stats per client, but getting empty list without
    any numbers
    >
    > I even run dd on client against mounted ceph fs, but output is
    only
    > like this:
    >
    > #> ceph fs perf stats 0 4638 192.168.121.1
    >
    > {"version": 2, "global_counters": ["cap_hit", "read_latency",
    > "write_latency", "metadata_latency", "dentry_lease",
    "opened_files",
    > "pinned_icaps", "
    > opened_inodes", "read_io_sizes", "write_io_sizes",
    "avg_read_latency",
    > "stdev_read_latency", "avg_write_latency", "stdev_write_latency",
    > "avg_metada
    > ta_latency", "stdev_metadata_latency"], "counters": [],
    > "client_metadata": {}, "global_metrics": {}, "metrics":
    > {"delayed_ranks": []}}
    >
    > Do I need to set some extra options?
    >
    Were you using the ceph-fuse/libcephfs user space clients ? if so
    you
    need to manually enable the 'client_collect_and_send_global_metrics'
    option, which is disabled in Ceph Pacific 16.2.13.

    While if you were using the kclient you need to make sure the
    'disable_send_metrics' module parameter is 'false' for ceph.ko
    module.

    Thanks

    - Xiubo


    > Does it work for some of you guys?
    >
    > Thank you
    >
    > dp
    > _______________________________________________
    > ceph-users mailing list -- ceph-users@xxxxxxx
    > To unsubscribe send an email to ceph-users-leave@xxxxxxx
    >
    _______________________________________________
    ceph-users mailing list -- ceph-users@xxxxxxx
    To unsubscribe send an email to ceph-users-leave@xxxxxxx

_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx




[Index of Archives]     [Information on CEPH]     [Linux Filesystem Development]     [Ceph Development]     [Ceph Large]     [Ceph Dev]     [Linux USB Development]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [xfs]


  Powered by Linux