[LU-15882] Lustre 2.15 GPUDirect Testing IO failed

Tue, 24 May 2022 05:53:40 +0000

gds sanity test found failed I/O:

/usr/local/gds/tools/gdsio -V -D /data/sanity/tests// -d 0  -w 8 -s 1G -i 32K:1024K:1K -x 0 -I 1 -o 1
Error: IO failed stopping traffic, fd :51 ret:-5 errno :1
Error: IO failed stopping traffic, fd :49 ret:-5 errno :1
Error: IO failed stopping traffic, fd :52 ret:-5 errno :1
Error: IO failed stopping traffic, fd :56 ret:-5 errno :1
Error: IO failed stopping traffic, fd :54 ret:-5 errno :1
Error: IO failed stopping traffic, fd :55 ret:-5 errno :1
FAILED

dmesg:

[Mon May 23 15:43:05 2022] nvidia-fs:write unable to flush dirty pages :-5
[Mon May 23 15:43:05 2022] nvidia-fs:write unable to flush dirty pages :-5
[Mon May 23 15:43:05 2022] nvidia-fs:write unable to flush dirty pages :-5
[Mon May 23 15:43:05 2022] nvidia-fs:write unable to flush dirty pages :-5
[Mon May 23 15:43:05 2022] nvidia-fs:write unable to flush dirty pages :-5
[Mon May 23 15:43:05 2022] nvidia-fs:write unable to flush dirty pages :-5

**************************************************
Testsuite : 181 / 182 tests passed
done tests:Mon May 23 22:02:56 UTC 2022

ddn@a100-01:~$ lctl get_param version
version=2.15.50_13_gc524079_dirty

NVIDIA-SMI 515.43.04

Driver Version: 515.43.04

CUDA Version: 11.7

Kernel: 5.4.0-109-generic

Whamcloud Community JIRA

[LU-15882] Lustre 2.15 GPUDirect Testing IO failed