1

Trying to find which rbd image is making most write-iops, but can't make any sense out from "rbd perf" output compared to "ceph status". What is difference between iops counters of ceph (160 op/s) vs rbd pref (WR 1/s)?

ceph status | grep client.
client: 493 KiB/s rd, 2.4 MiB/s wr, 10 op/s rd, 160 op/s wr

rbd perf image iostat.
NAME                      WR RD WR_BYTES RD_BYTES WR_LAT RD_LAT
ceph/vm-152-disk-0 1/s   0/s   71 KiB/s     0 B/s   13.04 ms   0.00 ns
ceph/vm-136-disk-0 0/s   0/s   819 B/s       0 B/s   919.79 us   0.00 ns

Mikko
  • 21
  • 4

1 Answers1

0

ceph status is summing io's for all pools.

As your rbd images are on the pool 'ceph', you can run 'ceph osd pool stats ceph' to get specific stats for that pool.

If you have only 1 WR/s on ceph/vm-152-disk-0 and 160 op/s wr on the whole cluster, it means that 159 op/s wr are done elsewhere, in another pool.

Francois Scheurer
  • 1
  • 2