fio fails to run more than 4 rbd jobs at once

[adippl]

Please acknowledge the following before creating a ticket

• [ x ] I have read the GitHub issues section of REPORTING-BUGS.

Description of the bug: fio fails to benchmark more than 4 RBD image jobs at once. There is no IO on ceph cluster and fio never finishes the benchmark

Environment: OS: ubuntu 22.04 librbd 18.2.2

fio version: 3.34 3.36. 3.37

Reproduction steps Example of my config file. Fio fails when running it. Commenting out one of the jobs and running with only 4 jobs makes fio run correctly. Moving ioengine= , clientname= or pool= makes no difference.

[global]
rw=randwrite
bs=4k
iodepth=16
group_reporting
time_based
ramp_time=30
runtime=60

[node00_image00]
ioengine=rbd
rbdname=node00_image00
clientname=benchmark
pool=benchmark_ssd

[node00_image01]
ioengine=rbd
rbdname=node00_image01
clientname=benchmark
pool=benchmark_ssd

[node00_image02]
ioengine=rbd
rbdname=node00_image02
clientname=benchmark
pool=benchmark_ssd

[node00_image03]
ioengine=rbd
rbdname=node00_image03
clientname=benchmark
pool=benchmark_ssd

[node00_image04]
ioengine=rbd
rbdname=node00_image04
clientname=benchmark
pool=benchmark_ssd

I've saved output of fio with --debug=all It's very long. I've uploaded it as a gist. https://gist.github.com/adippl/776dc4eda30e117688e2807d2e025e5e