Re: "ceph fs perf stats" and "cephfs-top" don't work

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

 



Contradicting outputs, it seems immediate ones. You need to wait for a few seconds for the 'perf stats' (and of-course the cephfs-top) to display the correct the metrics. I hope you have ongoing IO while running 'perf stats' and cephfs-top. What is your kernel version?

On 15/07/21 5:17 pm, Eugen Block wrote:
Hi,

I just setup a virtual one-node cluster (16.2.5) to check out cephfs-top. Regarding the number of clients I was a little surprised, too, in the first couple of minutes the number switched back and forth between 0 and 1 although I had not connected any client yet. But after a while the number got stable and correct. I have two clients connected now, but I don't see any stats despite having the stats module enabled:

---snip---
cephfs-top - Thu Jul 15 13:35:41 2021
Client(s): 2 - 0 FUSE, 0 kclient, 2 libcephfs

  client_id mount_root chit(%) rlat(s) wlat(s) mlat(s) dlease(%) ofiles oicaps oinodes mount_point@host/addr   24835     /          N/A     N/A     N/A     N/A     N/A N/A    N/A    N/A     N/A@pacific/v1:192.168.124.35   24846     /client    N/A     N/A     N/A     N/A     N/A N/A    N/A    N/A     N/A@host-192-168-124-168/v1:192.168.124.168
---snip---


The command 'ceph fs perf stats' also only shows this:

---snip---
pacific:~ # ceph fs perf stats
{"version": 1, "global_counters": ["cap_hit", "read_latency", "write_latency", "metadata_latency", "dentry_lease", "opened_files", "pinned_icaps", "opened_inodes"], "counters": [], "client_metadata": {"client.24835": {"IP": "v1:192.168.124.35", "hostname": "pacific", "root": "/", "mount_point": "N/A"}, "client.24846": {"IP": "v1:192.168.124.168", "hostname": "host-192-168-124-168", "root": "/client", "mount_point": "N/A"}}, "global_metrics": {"client.24835": [[0, 0], [0, 0], [0, 0], [0, 0], [0, 0], [0, 0], [0, 0], [0, 0]], "client.24846": [[0, 0], [0, 0], [0, 0], [0, 0], [0, 0], [0, 0], [0, 0], [0, 0]]}, "metrics": {"delayed_ranks": [], "mds.0": {"client.24835": [], "client.24846": []}}}
---snip---

although I have written a couple of GB into the cephfs.

Regards,
Eugen


Zitat von Erwin Bogaard <erwin.bogaard@xxxxxxxxx>:

Hi,

I just upgraded our cluster to pacific 16.2.5.
As I'm curious what cephfs-top could give for insights, I followed the
steps in the documentation.
After enabling the mgr module "stats":

# ceph mgr module ls
...
    "enabled_modules": [
        "dashboard",
        "iostat",
        "restful",
        "stats",
        "zabbix"
...

I tried the following command:
# ceph fs perf stats
{"version": 1, "global_counters": ["cap_hit", "read_latency",
"write_latency", "metadata_latency", "dentry_lease", "opened_files",
"pinned_icaps", "opened_inodes"], "counters": [], "client_metadata": {},
"global_metrics": {}, "metrics": {"delayed_ranks": []}}

As you can see, this returns no info whatsoever. The same with:

# cephfs-top
cluster ceph does not exist

The actual cluster name is "ceph".

So I don't understand why "ceph fs perf stats" isn't showing any
information.
Maybe another indicator something isn't ritght:

# ceph fs status
cephfs - 0 clients
======
RANK  STATE      MDS        ACTIVITY     DNS    INOS   DIRS CAPS
...

I see "0 clients". When I take a look in the mgr dashboard, I can actually
see all clients. Which are RHEL 7 & 8 cephfs kernel clients.
There is only 1 mds active, and 1 in standby-replay.
I have multiple pools active, but only 1 fs.

Does anyone have a suggestion where I can take a look enable gathering the
stats?
_______________________________________________
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