[LU-10225] sanity test 1 error: apply rmdir/rm on striped dir failed

• Type: Bug
• Resolution: Unresolved
• Priority: Critical
• Affects Version/s: Lustre 2.10.1
• Environment: Lustre2.10.1

Epic/Theme

Lustre-2.10.1 test zfs

• Severity: 4
• Assignee: WC Triage
• Reporter: fengyubo (Inactive)

Description

when run lustre with 2 MDS（active/inactive）share one mdt, the test below can pass.

when run with 2MDS（active/active）, the test get error shown below broken image link the corresponding script :

test_1() {
      test_mkdir -p $DIR/$tdir || error "mkdir $tdir failed"
      test_mkdir -p $DIR/$tdir/d2 || error "mkdir $tdir/d2 failed"
      test_mkdir $DIR/$tdir/d2 && error "we expect EEXIST, but not returned"
      $CHECKSTAT -t dir $DIR/$tdir/d2 || error "$tdir/d2 is not a dir"
      rmdir $DIR/$tdir/d2
      rmdir $DIR/$tdir
      $CHECKSTAT -a $DIR/$tdir || error "$tdir was not removed"
}

if MDSCOUNT in configuration file >1 ,the function test_mkdir will create a strped dir. 

Issue Links

• is related to: LU-3308 large readdir chunk size slows unlink/"rm -r" performance

Comments

[Andreas Dilger]

Is there something unusual with your test configuration? This test is running in DNE mode on our test clusters for several years without problems, since patch v2_3_57_0-56-g13b269a "LU-1187 tests: Add test_mkdir in sanity for DNE" in 2012. I changed this in patch v2_10_53_0-8-gc75aa6c "LU-3308 tests: fix sanity/sanityn test_mkdir() usage" to use "mkdir" instead of "test_mkdir" for this last directory, but it isn't clear why I did that. Getting the error output, and logs from the client and MDS would definitely help.

[James Nunez (Inactive)]

Would you please copy the error message(s) here and upload any logs that you have for this failure?