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
