--warnings-fatal swallows the error

[jlevon]

Please acknowledge the following before creating a ticket

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

Description of the bug:

Compare

[john.levon@john-levon-1 ansible]$ ../benchmark/fio/fio --client=10.117.29.229 --warnings-fatal ./jobfile_sequential_conditioning 
hostname=localhost.localdomain, be=0, 64-bit, os=Linux, arch=x86-64, fio=fio-3.39, flags=1
client: host=10.117.29.229 disconnected

with

[john.levon@john-levon-1 ansible]$ ../benchmark/fio/fio --client=10.117.29.229 ./jobfile_sequential_conditioning 
hostname=localhost.localdomain, be=0, 64-bit, os=Linux, arch=x86-64, fio=fio-3.39, flags=1
<localhost.localdomain> fio: time_based requires a runtime/timeout setting
<localhost.localdomain> sequential_conditioning: (g=0): rw=write, bs=(R) 65.5kB-65.5kB, (W) 65.5kB-65.5kB, (T) 65.5kB-65.5kB, ioengine=libaio, iodepth=64
<localhost.localdomain> Starting 1 thread

There appears to be no way to get the actual warning text above when specifying --warnings-fatal

Environment: client side: Centos 7; server side: Centos 8 Stream

fio version: fio-3.39

Reproduction steps

jobfile like this will be sufficient to demonstrate:

[sequential_conditioning]
filename=/dev/nvme0n1
readwrite=write
time_based=1

[dakralex]

I couldn't reproduce this locally with fio-3.39 on Arch, are client and server both at the same version? Does the --warnings-fatal show the error correctly when run locally on the server side?

% fio --client=10.0.0.2 ./testjob
hostname=localhost.localdomain, be=0, 64-bit, os=Linux, arch=x86-64, fio=fio-3.39, flags=1
<localhost.localdomain> fio: time_based requires a runtime/timeout setting
<localhost.localdomain> sequential_conditioning: (g=0): rw=write, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=psync, iodepth=1
<localhost.localdomain> Starting 1 thread

% fio --client=10.0.0.2 --warnings-fatal ./testjob
hostname=localhost.localdomain, be=0, 64-bit, os=Linux, arch=x86-64, fio=fio-3.39, flags=1
<localhost.localdomain> fio: time_based requires a runtime/timeout setting

[jlevon]

They are both at the same version, commit a6e474c

Server side:

[root@localhost fio]# ./fio --warnings-fatal job
fio: time_based requires a runtime/timeout setting

[jlevon]

I see the same with current head on both sides, 84787ad

[vincentkfu]

I can't reproduce this:

vincent@localhost:~/fio-dev/1977$ cat > 1977.fio
[sequential_conditioning]
filename=/dev/nvme0n1
readwrite=write
time_based=1
vincent@localhost:~/fio-dev/1977$ ../fio-canonical/fio --client=localhost 1977.fio
hostname=localhost, be=0, 64-bit, os=Linux, arch=x86-64, fio=fio-3.41-5-g84787, flags=1
<localhost> fio: time_based requires a runtime/timeout setting
<localhost> sequential_conditioning: (g=0): rw=write, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=psync, iodepth=1
<localhost> Starting 1 process
<localhost> sequential_conditioning: you need to specify size=
<localhost> fio: pid=0, err=22/file:filesetup.c:1151, func=total_file_size, error=Invalid argument
<localhost>
client <localhost>: exited with error 1

Run status group 0 (all jobs):
vincent@localhost:~/fio-dev/1977$ ../fio-canonical/fio --client=localhost --warnings-fatal 1977.fio
hostname=localhost, be=0, 64-bit, os=Linux, arch=x86-64, fio=fio-3.41-5-g84787, flags=1
<localhost> fio: time_based requires a runtime/timeout setting

You could run with --debug=net to see what is being passed between the client and the server.