[LU-5842] sanity-sec tests 13 and 14 use a lot more time than necessary

• Status: Resolved
• Project: Lustre
• Fix Version/s: Lustre 2.7.0
• Type: Bug
• Priority: Minor
• Reporter: Oleg Drokin
• Assignee: WC Triage
• Resolution: Fixed
• Severity: 3

Description

It looks like tests 14 and 14 of sanity-sec.sh do neste loops 3 levels long: 16 x 3 x 256 and as such it takes 2600+ seconds each.
Sample run: https://testing.hpdd.intel.com/test_sets/ec53a19c-6144-11e4-a66b-5254006e85c2

The code:

        for ((i = 0; i < NODEMAP_COUNT; i++)); do
                for ((j = 0; j < NODEMAP_RANGE_COUNT; j++)); do
                        for ((k = 1; k < 253; k++)); do
                                if ! test_nid $SUBNET_CHECKSUM.$i.$j.$k \
                                       ${HOSTNAME_CHECKSUM}_${i}; then
                                        rc=$((rc + 1))
                                fi
                        done
                done
        done

This last loop could easily be transferred into:

for k in 1 10 64 128 200 250 ; do

for a great reduction in runtime and at the same - to space out the checks instead of checkign just some initial values
(as apparently was envisioned by now unused NODEMAP_IPADDR_COUNT variable that could be removed or replaced by NODEMAP_IPADDR_LIST to hold the values above instead for easy variability).

Issue Links

• is related to: LU-5847 sanity-sec: lctl nodemap_test_nid on ...

Comments

[Andreas Dilger]

Kit, we were discussing similar issues with the tests about to be added in you other nodemap testing patch.

[Kit Westneat]

Ok, should incorporate these changes into the new test patch?
http://review.whamcloud.com/#/c/10406/

Or would it be better to create a new small patch for quicker landing?

[Andreas Dilger]

Probably better to have separate patches. No point to entangle two issues if they can land separately.

[Kit Westneat]

Ok, I put up a new patch here:
http://review.whamcloud.com/#/c/12532/

[Jodi Levi (Inactive)]

Patch landed to Master.