[LU-12146] remove "skip_env" for FSTYPE=zfs Created: 02/Apr/19  Updated: 29/Jul/19

Status: Open
Project: Lustre
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Improvement Priority: Minor
Reporter: Andreas Dilger Assignee: WC Triage
Resolution: Unresolved Votes: 0
Labels: test_script_improvements, tests

Issue Links:
Related
is related to LU-10461 sanity 77c failure due to variables c... Resolved
Rank (Obsolete): 9223372036854775807

 Description   

Tests should not use skip_env if they are skipped because FSTYPE != "ldiskfs".

Such tests would always be skipped when the filesystem type is ZFS, and this is not an indication of a deficient test environment. The skip_env() check is meant for detecting test environments that are lacking things like test programs (e.g. dbench or IOR) or are configured incorrectly (e.g. MDT or OST is too small, or not enough of them for basic tests). The intent is that skip_env() allows such tests to be skipped during developer testing, where the test environment may not have everything configured, but it should be considered an error() if the autotest environment does not have everything installed properly in order to execute the test case.

The regular "skip" check is intended to skip tests when the configuration is unsuitable for the current test for whatever reason, so we don't need to mark such cases with "skip_env".

It looks like several such checks are in the current tests:

  • sanity.sh test 17m (ldiskfs), 17n, 24E, 51e, 57a, 57b, 60a, 77a (GSS), 77b, 77c, 77d, 77f, 77g, 77k, 100, 101a, 103a, 129, 130a (ZFS), 130b, 130c, 130d, 130e, 228a, 228b, 228c, 233a, 233b, 252, 315, 399b, 804,


 Comments   
Comment by Andreas Dilger [ 29/Jul/19 ]

There are a bunch of other places that skip_env is incorrectly used, in particular for $MDSCOUNT and other places that are just variations in the test configs, and not shortcomings in the Lustre autotest environment. I guess we could force all tests to run with multiple MDTs, but I don't think that is the case today.

It would be interesting to at least try one test run with $FAIL_ON_SKIP_ENV=true and see what fails. Ideally once these patches are run and any resulting failures are fixed, then we could enable this by default in the future?

Generated at Sat Feb 10 02:50:03 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.