Jian,
the sanity test_130* failures on SLES15 look to be the same as LU-10335, which is a bug in upstream e2fsprogs until 1.45 and 1.44.2, according to comments in that ticket.
Looking at the output of your test log, it shows that the SLES15 SP1 client is using e2fsprogs-1.43.8:
This problem is hit if users on SLES15 clients run "filefrag -v" on a Lustre file (or any other file with unknown FIEMAP flags) with the distro e2fsprogs. This is fairly unlikely, but not impossible. If they run with the Lustre-patched e2fsprogs on the client (based on 1.45.2) this problem is already fixed.
Neil,
I'm not sure whether an issue like this should be fixed in the SLES15 e2fsprogs-1.43.8 or not? The original patch is at https://marc.info/?l=linux-ext4&m=151993561511549 or upstream e2fsprogs commit v1.44.1-1-g17a1f2c1 for reference. Of course, updating to e2fsprogs-1.45.6 would be even better than applying just this one fix, but I'm not sure of your policy for mid-series updates.
In any case, this ticket can be closed because the Lustre e2fsprogs are building and passing their internal tests on SLES15, and have shown proper functionality for everything besides this one issue.
Jian,
the sanity test_130* failures on SLES15 look to be the same as
LU-10335, which is a bug in upstream e2fsprogs until 1.45 and 1.44.2, according to comments in that ticket.Looking at the output of your test log, it shows that the SLES15 SP1 client is using e2fsprogs-1.43.8:
This problem is hit if users on SLES15 clients run "filefrag -v" on a Lustre file (or any other file with unknown FIEMAP flags) with the distro e2fsprogs. This is fairly unlikely, but not impossible. If they run with the Lustre-patched e2fsprogs on the client (based on 1.45.2) this problem is already fixed.
Neil,
I'm not sure whether an issue like this should be fixed in the SLES15 e2fsprogs-1.43.8 or not? The original patch is at https://marc.info/?l=linux-ext4&m=151993561511549 or upstream e2fsprogs commit v1.44.1-1-g17a1f2c1 for reference. Of course, updating to e2fsprogs-1.45.6 would be even better than applying just this one fix, but I'm not sure of your policy for mid-series updates.
In any case, this ticket can be closed because the Lustre e2fsprogs are building and passing their internal tests on SLES15, and have shown proper functionality for everything besides this one issue.