Details
-
Bug
-
Resolution: Unresolved
-
Minor
-
None
-
None
-
None
-
3
-
9223372036854775807
Description
This issue was created by maloo for sarah <sarah@whamcloud.com>
This issue relates to the following test suite run: https://testing.whamcloud.com/test_sets/49ab1da6-29b3-4190-8824-1118bce3af5b
test_27D failed with the following error:
llapi_layout_test failed
Test session details:
clients: https://build.whamcloud.com/job/lustre-reviews/102975 - 4.18.0-477.27.1.el8_8.x86_64
servers: https://build.whamcloud.com/job/lustre-reviews/102975 - 4.18.0-477.27.1.el8_lustre.x86_64
llapi_layout_test test16 and 20 failed. it is because the "lfs getstripe -c" return the last component's stripe count for dir, but return the actual used stripe count for file, so the value doesn't match in test 16 and 20
Per Andreas's suggestion, we should improve llapi_layout_test to work with PFL layout
since that subtest is actually testing llapi_layout_*() routines, my preference would be to update the failing test cases to understand if the source is a PFL layout and handle it properly, rather than skip the subtests. That way, we get more testing that llapi_layout itself is working rather than less. Andreas, 6:06 PM like you wrote earlier, there is a bit of a gap between LAYOUT_GET_EXPECTED from the parent directory layout (returns last stripe), and what the actual 0-length file is using (returns last used) stripe. It might be enough to truncate() the PFL file up to the last component (size=last component start + 64KB) so that all the objects are allocated. that would work for a regular layout (last component starts at offset 0) as well as a PFL layout
VVVVVVV DO NOT REMOVE LINES BELOW, Added by Maloo for auto-association VVVVVVV
sanity-compr test_27D - llapi_layout_test failed
