[mesos] branch 1.9.x updated: Fixed handling disconnected agents by REACTIVATE_AGENT call.

[asekretenko]

This is an automated email from the ASF dual-hosted git repository.

asekretenko pushed a commit to branch 1.9.x
in repository https://gitbox.apache.org/repos/asf/mesos.git


The following commit(s) were added to refs/heads/1.9.x by this push:
     new 70291ed  Fixed handling disconnected agents by REACTIVATE_AGENT call.
70291ed is described below

commit 70291edf09f5b35af2b5389024de84b550ccacf3
Author: Andrei Sekretenko <asekrete...@apache.org>
AuthorDate: Tue Apr 14 20:05:11 2020 +0200

    Fixed handling disconnected agents by REACTIVATE_AGENT call.
    
    This patch fixes MESOS-10116 by preventing REACTIVATE_AGENT from
    activating disconnected agents in the allocator and also fixes the
    handling of agents that were removed while the reactivation was being
    stored into the registry.
    
    Review: https://reviews.apache.org/r/72363
---
 src/master/http.cpp   | 14 ++++++++++++--
 src/master/master.cpp | 17 ++++-------------
 src/master/master.hpp |  4 ----
 3 files changed, 16 insertions(+), 19 deletions(-)

diff --git a/src/master/http.cpp b/src/master/http.cpp
index 9cfdc77..db456b5 100644
--- a/src/master/http.cpp
+++ b/src/master/http.cpp
@@ -4145,8 +4145,18 @@ Future<Response> Master::Http::_reactivateAgent(
       master->slaves.deactivated.erase(slaveId);
 
       Slave* slave = master->slaves.registered.get(slaveId);
-      if (slave != nullptr) {
-        master->reactivate(slave);
+      if (slave == nullptr) {
+        return Conflict("Agent removed while processing the call");
+      }
+
+      if (slave->connected) {
+        LOG(INFO) << "Reactivating agent " << *slave;
+
+        slave->active = true;
+        master->allocator->activateSlave(slaveId);
+      } else {
+        LOG(INFO) << "Disconnected agent " << *slave
+                  << " will be reactivated upon reregistration.";
       }
 
       slave->estimatedDrainStartTime = None();
diff --git a/src/master/master.cpp b/src/master/master.cpp
index a99f0eb..f641adc 100644
--- a/src/master/master.cpp
+++ b/src/master/master.cpp
@@ -3447,18 +3447,6 @@ void Master::deactivate(Slave* slave)
 }
 
 
-void Master::reactivate(Slave* slave)
-{
-  CHECK_NOTNULL(slave);
-  CHECK(!slaves.deactivated.contains(slave->id));
-
-  LOG(INFO) << "Reactivating agent " << *slave;
-
-  slave->active = true;
-  allocator->activateSlave(slave->id);
-}
-
-
 void Master::resourceRequest(
     const UPID& from,
     const FrameworkID& frameworkId,
@@ -8173,7 +8161,10 @@ void Master::___reregisterSlave(
     dispatch(slave->observer, &SlaveObserver::reconnect);
 
     if (!slaves.deactivated.contains(slave->id)) {
-      reactivate(slave);
+      LOG(INFO) << "Reactivating re-registered agent " << *slave;
+
+      slave->active = true;
+      allocator->activateSlave(slave->id);
     }
   }
 
diff --git a/src/master/master.hpp b/src/master/master.hpp
index c4a1e62..7dc6248 100644
--- a/src/master/master.hpp
+++ b/src/master/master.hpp
@@ -685,10 +685,6 @@ protected:
   // offers). Other aspects of the agent will continue to function normally.
   void deactivate(Slave* slave);
 
-  // Adds the agent back to the resource offer cycle.
-  // Must *NOT* be called if the agent is `deactivated`.
-  void reactivate(Slave* slave);
-
   // Add a slave.
   void addSlave(
       Slave* slave,