-
Type: Bug
-
Resolution: Fixed
-
Priority: Major - P3
-
Affects Version/s: None
-
Component/s: Replication
-
Fully Compatible
-
ALL
-
v3.6, v3.4
-
46
PingStats, defined in ToplogyCoordinatorImpl, is a data structure used to track statistics about replication heartbeat messages. It has a variable called _numFailuresSinceLastStart, which keeps track of failed heartbeat attempts. When a good heartbeat response comes back, this variable is set to std::numeric_limits<int>::max(), presumably as an easy way for this value to always be larger than others in certain comparisons. In the case that we actually get back a good heartbeat response (PingStats::hit) and then a failed heartbeat from the same node (PingStats::miss), we will actually overflow this _numFailuresSinceLastStart variable. We should not use the integer max as a way to determine that we received a good heartbeat response. Creating a separate field in PingStats to explicitly track this would probably be a simple approach.
- links to