[LU-3332] sanity-scrub and sanity-lfsck need to support DNE

• Type: Bug
• Resolution: Fixed
• Priority: Blocker
• Fix Version/s: Lustre 2.5.0
• Affects Version/s: Lustre 2.5.0
• Labels: MB
• Severity: 3
• Assignee: Li Wei (Inactive)
• Reporter: Di Wang

Description

Currently, sanity-scrub.sh still use SINGLEMDS in the script, which make it unable to run with DNE configuration.

Issue Links

• is duplicated by: LU-3150 Add dne tests in sanity-lfsck and sanity-scrub
• is related to: LU-3420 OI scrubbing could not automatically engage after restoring a secondary MDT from a (file-level) backup
• is related to: LU-2915 LFSCK 1.5 technical debts

Comments

[Jodi Levi (Inactive)]

Patch has landed to Master. Please let me know if more work is needed in this ticket and I will reopen.

[Li Wei (Inactive)]

I can't imagine how happy I'd be if I knew you were to on this. Too bad neither of us had left any obvious messages here earlier.

[Keith Mannthey (Inactive)]

Thanks for the link to that patch. Peter had asked me to look at this LU earlier this week.

That is exactly the style of patch I was working on and will give it a good review.

[Di Wang]

Hmm, I thought Li wei worked on this ticket already. http://review.whamcloud.com/#change,6333

[Keith Mannthey (Inactive)]

Di Wang is correct a great many changes are needed. Not quite every line will have to be touched touched but pretty close.

I am a bit of a ways into a rewrite for sanity-scrub and so far a few spots in test-framework have also be identified.

There seems to be some loose timing built into this test in general. Spots where you expect a given transition state, I am not sure if these time checks will work as well with DNE. I am not going to focus on Logic changes to the test at this point just a conversion to DNE safe calls for v1.

I think an incremental approach to these changes keep the reviewing of all of this sane, the removal of SINGLEMDS is no small task even just in this one test. I will work on getting some form of a reviewable patch into the system tomorrow that is not a totally a for_test_only sort of patch and setup a proper expect always to allow DNE to run a few tests.

My approach is slightly different that Di Wangs so I will open a new patch or two. I like use of helper functions to make the test more readable and radically decrease the amount of code.

for instance my test_0 presently looks like:

test_0() {
        local num

        scrub_prep 0
        echo "start $SINGLEMDS without disabling OI scrub"

        start_all_mds($MOUNT_OPTS_SCRUB)

        check_status("init")
        check_flags()

        mount_client $MOUNT || error "Failed to start client!"
        check_diff()
}
run_test 0 "Do not auto trigger OI scrub for non-backup/restore case"

[Di Wang]

http://review.whamcloud.com/6333 Unfortunately, it might need many changes, and I do not have time to finish all of them. I only fix 0 and 1a. Maybe someone else can help to fix all of the script problem. So sanity-scrub can be run with DNE. Thanks.