tsan flake in xds_fuzz

[alyssawilk]

https://dev.azure.com/cncf/4684fb3d-0389-4e0b-8251-221942316e06/_apis/build/builds/45949/logs/90

CorpusExamples/FuzzerCorpusTest.RunOneCorpusFile/6

2020-07-23T18:20:32.8483285Z ==================
2020-07-23T18:20:32.8483632Z WARNING: ThreadSanitizer: data race (pid=15)
followed by thousands of lines of tsan gorp.

I've often found when tsan is noisy this way it's because some underlying expectation has failed and the test is going off in an unexpected direction. There was one failure of this type which might help debug?

() const
2020-07-23T18:20:33.0312638Z ==================
2020-07-23T18:20:33.0313102Z test/server/config_validation/xds_fuzz.cc:344: Failure
2020-07-23T18:20:33.0313477Z Expected equality of these values:
2020-07-23T18:20:33.0313769Z routes.size()
2020-07-23T18:20:33.0314041Z Which is: 2
2020-07-23T18:20:33.0314302Z dump.size()
2020-07-23T18:20:33.0314549Z Which is: 1
2020-07-23T18:20:33.0314815Z Stack trace:
2020-07-23T18:20:33.0315476Z 0x29dcae8: Envoy::XdsFuzzTest::verifyRoutes()
2020-07-23T18:20:33.0315827Z 0x29da58c: Envoy::XdsFuzzTest::verifyState()
2020-07-23T18:20:33.0316270Z 0x29da3e8: Envoy::XdsFuzzTest::replay()
2020-07-23T18:20:33.0316597Z 0x292cc26: Envoy::TestOneProtoInput()
2020-07-23T18:20:33.0316917Z 0x292c7b1: LLVMFuzzerTestOneInput
2020-07-23T18:20:33.0317429Z 0x53d03ef: Envoy::(anonymous namespace)::FuzzerCorpusTest_RunOneCorpusFile_Test::TestBody()
2020-07-23T18:20:33.0317908Z 0x7b311fd: testing::internal::HandleSehExceptionsInMethodIfSupported<>()
2020-07-23T18:20:33.0318346Z 0x7b151df: testing::internal::HandleExceptionsInMethodIfSupported<>()
2020-07-23T18:20:33.0318726Z 0x7afa432: testing::Test::Run()
2020-07-23T18:20:33.0319345Z 0x7afb2d4: testing::TestInfo::Run()
2020-07-23T18:20:33.0319672Z ... Google Test internal frames ...

[htuch]

Another example in https://dev.azure.com/cncf/envoy/_build/results?buildId=45948&view=logs&j=5eb754ee-449a-545e-2b62-07fe525d7cbc&t=1110550d-5f5f-55a9-69fd-d02adf65e59a. @samflattery can you take a look?

[mattklein123]

@htuch this is flaking often enough that if @samflattery doesn't have time to look today I would suggest we disable for right now. Will defer to both of you on fix timing.

[htuch]

I agree; since this is a fuzzer we can either disable or rollback without impact.

[samflattery]

Just seeing this now. I'll take a look and will hopefully push a fix soon

[samflattery]

Should be fixed with #12246