[jira] (MESOS-7036) Rate limiter deadlocks during IO Switchboard-related tests

[JIRA]

Title: Message Title

Gastón Kleiman commented on MESOS-7036   Re: Rate limiter deadlocks during IO Switchboard-related tests I observed the same while working on https://reviews.apache.org/r/55901/. The deadlock is triggered by the loop in the following lines: https://github.com/apache/mesos/blob/d0f4cd4d9e98c27fd03eeafcef6e16bb04f51bc8/src/slave/containerizer/mesos/io/switchboard.cpp#L682-L706 What I observed is that RateLimiter::~RateLimiter is sometimes invoked by the thread that is executing RateLimiterProcess::_acquire . The destructor calls wait(limiterProcess), which triggers the deadlock. I managed to get rid of the deadlock by applying the following diff to line 696: --- i/src/slave/containerizer/mesos/io/switchboard.cpp +++ w/src/slave/containerizer/mesos/io/switchboard.cpp @@ -693,7 +693,7 @@ Future<http::Connection> IOSwitchboard::connect( } return Break(); }) - .then(defer(self(), [=]() -> Future<http::Connection> { + .then(defer(self(), [this, limiter, address, containerId]() -> Future<http::Connection> { if (!infos.contains(containerId)) { return Failure("I/O switchboard has shutdown"); } I think that this is a brittle "fix", so I am trying to understand the loop better, in order to propose a better solution. Add Comment   This message was sent by Atlassian JIRA (v6.3.15#6346-sha1:dbc023d)