Patrick Rhomberg created GEODE-5299:
---------------------------------------
Summary: Correct failure resulting from potential race condition
in geode-wan:CreateRegionCommandDUnitTest
Key: GEODE-5299
URL: https://issues.apache.org/jira/browse/GEODE-5299
Project: Geode
Issue Type: Test
Reporter: Patrick Rhomberg
A recent commit contains the line the effective sequence:
{noformat}
locator = lsRule.startLocatorVM(0);
server1 = lsRule.startServerVM(1, locator.getPort());
server2 = lsRule.startServerVM(2, locator.getPort());
gfsh.connectAndVerify(locator); String regionName =
testName.getMethodName();
String gatewaySenderName = "gatewaySender";
IgnoredException.addIgnoredException("could not get remote locator
information");
gfsh.executeAndAssertThat(
"create gateway-sender --parallel=true --remote-distributed-system-id=2
--id="
+ gatewaySenderName)
.statusIsSuccess();
locator.waitTilGatewaySendersAreReady(1);
{noformat}
Notably, the {{waitTilGatewaySendersAreReady}} expects to find 1 gateway mbean
from the two servers on which the gateway will be created.
However, the underlying implementation of the method expects to find *exactly*
1 bean, meaning that if the gateway creation and bean registration occurs
before the {{wait*}} method is called, the test will fail to a hang despite
correct execution.
The true root cause of this error is likely the ambiguity in the method and
variable naming. Consider renaming this method and related to be explicitly
clear to future developers that these methods have exact expectation.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
