Fixed the flaky test in the ProtectedPathSpec #5648
Conversation
|
Please see my comments and screenshots under #5608 issue |
Implements #5390 Related to #5608 * Fixed the test "Flow swaps to protected path when main path gets broken, becomes DEGRADED if protected path is unable to reroute(no bw)" * Earlier in some cases when the switchPair was set to 2-3 or 8-9, the protected path had only 1 ISL and it had enough BW, so the test failed because the BW was not reduced for some protected path ISLs. * Now this test passes even when the swPair is 2-3 or 8-9, and other switches. So the temporary fix to skip 8-9 switches is removed. * Removed the workaround to skip 8-9 from 2 tests
izadorozhna
force-pushed
the
test/5390_flaky_protected_path_spec
branch
from
b058639
to
7dc6659
Compare
| def switchPair = switchPairs.all() | ||
| .excludeSwitches(switchesWhere5608IsReproducible) | ||
| .withAtLeastNNonOverlappingPaths(2).random() | ||
| def switchPair = switchPairs.all().withAtLeastNNonOverlappingPaths(2).random() |
There was a problem hiding this comment.
What exactly is the fix? #5608 is still open and not solved, as far as I can see.
There was a problem hiding this comment.
Yes, let me add some details to explain my PR.
So, I was debugging a couple of tests (namely, "Flow swaps to protected path when main path gets broken, becomes DEGRADED if protected path is unable to reroute(no bw)" and "Flow swaps to protected path when main path gets broken, becomes DEGRADED if protected path is unable to reroute(no path)" because they failed sometimes for the past month. The failure was related to the same Couldn't find non overlapping protected path. Skipped creating it error message mentioned in #5608. So, even with the WA to skip 8-9 switches, the overlapping issue was still reproducible.
Flow swaps to protected path when main path gets broken, becomes DEGRADED if protected path is unable to reroute(no bw)- here the issue was still reproducible between the switches 2-3, please see the screenshot below.
When I investigated this test, it turned out that the protected path ISL has enough bw, that is why the test cannot get expected "Not enough bandwidth or no path found" message. So the fix in this PR is done to reduce the BW on the original protected path ISLs. So, with this fix, the test passes because the flow is degraded due to the expected "Not enough bandwidth or no path found" message. So, now no need to skip 8-9 switches since with the correct BW reducing, the test passes.Flow swaps to protected path when main path gets broken, becomes DEGRADED if protected path is unable to reroute(no path)- this test also failed sometimes due to Overlapping issue - even with the WA to skip 8-9 switches. I made the PR Fixed the flaky test when the flow swaps to protected path #5645 to fix this test because the ISLs were not broken correctly there. But now I see that with this fix from Fixed the flaky test when the flow swaps to protected path #5645, the WA to skip 8-9 can be removed, it is not reproduced anymore.
So, I have executed both these tests 20 times with just 8-9 switchpair, 2-3 switchpair, random switchpair and the overlapping issue is not reproduced anymore - the tests are passing. Thus, there is no need to exclude 8-9 switches from the switchpair anymore.
|
|
||
| when: "Create flow with protected path" | ||
| def flow = flowHelperV2.randomFlow(switchPair).tap { allocateProtectedPath = true } | ||
| flowHelperV2.addFlow(flow) | ||
| def path = northbound.getFlowPath(flow.flowId) | ||
|
|
||
| and: "Other paths have not enough bandwidth to host the flow in case of reroute" | ||
| def otherIsls = switchPair.paths.findAll { it != pathHelper.convert(path.protectedPath) && |
There was a problem hiding this comment.
when we removed path.protectedPath from exclusion, we set up bandwidth less that required (protected path), but after isl break, our new main path is UP (very strange behaviour). Need to discuss.
|
Closing since the fix was not correct. It was not correct to break the used ISLs in protected or main path. |
|
The new PR will be created instead |
Implements #5390
Related to #5608