[IMPALA-5904] Enable ThreadSanitizer for Impala

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: Infrastructure
• Labels: ramp-up
• Epic Color: ghx-label-4
• Assignee: Sahil Takiar
• Reporter: Tim Armstrong
• Created: 06/Sep/17 18:14
• Updated: 01/Sep/20 22:56

Description

It would be great to be able to automatically detect data races in Impala using ThreadSanitizer to avoid tricky-to-reproduce bugs. This issue tracks enabling ThreadSanitizer, fixing bugs and adding suppressions to get to the point where Impala runs cleanly with the sanitizer.

Issue Links

• is blocked by: IMPALA-4658 Potential race if compiler reorders ReachedLimit() usage
• is related to: IMPALA-9611 Hang in HandoffToProbesAndWait() for multithreaded join build
• relates to: IMPALA-9612 Runtime filter wait longer than it should be
• relates to: IMPALA-9840 ThreadSanitizer: data race internal-queue.h in InternalQueueBase::Enqueue
• relates to: IMPALA-10121 bin/jenkins/finalize.sh should generate JUnitXML for TSAN failures

Sub-Tasks

• Add ThreadSanitizer build: Resolved
• Add ThreadSanitizer to https://jenkins.impala.io/job/all-build-options/: Resolved
• TSAN data race in ReservationTracker::used_reservation_ updates: Resolved
• TSAN data race in TmpFileMgr::File::Blacklist: Resolved
• Possible data race in TupleDescriptor::GetLlvmStruct: Resolved
• Template tuples are initialized multiple times: Resolved
• TSAN data race in kudu::Socket::Accept and TraceLog::UpdateCategoryGroupEnabledFlag: Resolved
• TSAN data race in kudu::ScopedWatchKernelStack / KernelStackWatchdog: Resolved
• Instantiations/ExprTest.MathConversionFunctions fails in TSAN builds: Resolved
• Possible data race in kudu::security::InitKerberosForServer: Resolved
• TSAN data race in RuntimeFilterBank::CancelLocked(): Resolved
• TSAN data races during impalad shutdown: Resolved
• Set halt_on_error=1 for TSAN builds: Resolved
• TSAN data race in hdfs-bulk-ops: Resolved
• Data race on Webserver::UrlHandler::is_on_nav_bar_: Resolved
• Thrift server shutdown is racey: Open
• Allow TSAN to be set on codegen: Open
• Possible data race in BloomFilter::BucketFindAVX2: Open
• Compile thirdparty toolchain libs with TSAN: Open
• TSAN error in ImpalaServer::QueryStateRecord::Init: Resolved