private async Task ValidateClusterHealth(FabricTestContext testContext, ValidateClusterAction action, CancellationToken token)
{
TimeoutHelper timer = new TimeoutHelper(action.MaximumStabilizationTimeout);
bool success = false;
StringBuilder healthinfo = new StringBuilder();
while (!success && timer.GetRemainingTime() > TimeSpan.Zero)
{
healthinfo.Clear();
ClusterHealthPolicy healthPolicy = new ClusterHealthPolicy();
var clusterHealthResult = await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() =>
testContext.FabricClient.HealthManager.GetClusterHealthAsync(
healthPolicy,
action.RequestTimeout,
token),
FabricClientRetryErrors.GetEntityHealthFabricErrors.Value,
timer.GetRemainingTime(),
token).ConfigureAwait(false);
bool checkError = (action.CheckFlag & ValidationCheckFlag.CheckError) != 0;
bool checkWarning = (action.CheckFlag & ValidationCheckFlag.CheckWarning) != 0;
if ((checkError && clusterHealthResult.AggregatedHealthState == HealthState.Error) ||
(checkWarning && clusterHealthResult.AggregatedHealthState == HealthState.Warning) ||
clusterHealthResult.AggregatedHealthState == HealthState.Invalid ||
clusterHealthResult.AggregatedHealthState == HealthState.Unknown)
{
AppTrace.TraceSource.WriteInfo(TraceSource, "Cluster health state is {0}. Will Retry check", clusterHealthResult.AggregatedHealthState);
foreach (HealthEvent healthEvent in clusterHealthResult.HealthEvents)
{
healthinfo.AppendLine(string.Format(
"Cluster health state is '{0}' with property '{1}', sourceId '{2}' and description '{3}'",
healthEvent.HealthInformation.HealthState,
healthEvent.HealthInformation.Property,
healthEvent.HealthInformation.SourceId,
healthEvent.HealthInformation.Description));
}
AppTrace.TraceSource.WriteInfo(TraceSource, healthinfo.ToString());
}
else
{
success = true;
}
if (!success)
{
// Delay before querying again so we allow some time for state to change - don't spam the node
await AsyncWaiter.WaitAsync(RetryWaitTimeout, token).ConfigureAwait(false);
}
}
if (!success)
{
throw new FabricValidationException(StringHelper.Format(StringResources.Error_ServiceNotHealthy, "Cluster", action.MaximumStabilizationTimeout, healthinfo));
}
}
protected override async Task ExecuteActionAsync(FabricTestContext testContext, StartNodeAction action, CancellationToken cancellationToken)
{
ThrowIf.Null(action.NodeName, "NodeName");
this.helper = new TimeoutHelper(action.ActionTimeout);
string nodeName = action.NodeName;
BigInteger nodeInstance = action.NodeInstance;
if (nodeInstance == BigInteger.MinusOne)
{
var nodeInfo = await GetCurrentNodeInfoAsync(testContext, action, cancellationToken);
if (nodeInfo == null)
{
throw new FabricException(StringResources.Error_NodeNotFound, FabricErrorCode.NodeNotFound);
}
nodeInstance = nodeInfo.NodeInstanceId;
}
await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.FaultManager.StartNodeUsingNodeNameAsync(
nodeName,
nodeInstance,
action.IPAddressOrFQDN,
action.ClusterConnectionPort,
action.RequestTimeout, cancellationToken),
this.helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
if (action.CompletionMode == CompletionMode.Verify)
{
bool success = false;
while (this.helper.GetRemainingTime() > TimeSpan.Zero)
{
var nodeInfo = await this.GetCurrentNodeInfoAsync(testContext, action, cancellationToken).ConfigureAwait(false);
if (nodeInfo != null && nodeInfo.NodeInstanceId > nodeInstance && nodeInfo.IsNodeUp)
{
success = true;
break;
}
ActionTraceSource.WriteInfo(TraceSource, "NodeName = {0} not yet Started. Retrying...", action.NodeName);
await AsyncWaiter.WaitAsync(TimeSpan.FromSeconds(5), cancellationToken).ConfigureAwait(false);
}
if (!success)
{
throw new TimeoutException(StringHelper.Format(StringResources.Error_TestabilityActionTimeout,
"StartNode",
action.NodeName + ":" + action.NodeInstance));
}
}
action.Result = new StartNodeResult(action.NodeName, nodeInstance);
this.ResultTraceString = StringHelper.Format("StartNodeAction succeeded for {0}:{1} with CompletionMode = {2}", action.NodeName, nodeInstance, action.CompletionMode);
}
protected override async Task ExecuteActionAsync(FabricTestContext testContext, InvokeQuorumLossAction action, CancellationToken cancellationToken)
{
ThrowIf.Null(action.PartitionSelector, "PartitionSelector");
var helper = new TimeoutHelper(action.ActionTimeout);
// get info about the service so we can check type and trss
ServiceDescription result = await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.ServiceManager.GetServiceDescriptionAsync(
action.PartitionSelector.ServiceName,
action.RequestTimeout,
cancellationToken),
helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
if (result.Kind != ServiceDescriptionKind.Stateful)
{
throw new InvalidOperationException(StringHelper.Format(StringResources.Error_InvalidServiceTypeTestability, "QuorumLoss", "Stateful", action.PartitionSelector.ServiceName, "Stateless"));
}
StatefulServiceDescription statefulServiceDescription = result as StatefulServiceDescription;
ReleaseAssert.AssertIf(statefulServiceDescription == null, "Service is not a stateful service");
if (!statefulServiceDescription.HasPersistedState)
{
throw new InvalidOperationException(StringHelper.Format(StringResources.Error_InvalidServiceTypeTestability, "QuorumLoss", "Stateful Persistent", action.PartitionSelector.ServiceName, "Stateful In-Memory Only"));
}
// figure out /which/ partition to select
var getPartitionStateAction = new GetSelectedPartitionStateAction(action.PartitionSelector)
{
RequestTimeout = action.RequestTimeout,
ActionTimeout = helper.GetRemainingTime()
};
await testContext.ActionExecutor.RunAsync(getPartitionStateAction, cancellationToken);
Guid partitionId = getPartitionStateAction.Result.PartitionId;
// get data about replicas in that partition
ServiceReplicaList replicasResult = await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.QueryManager.GetReplicaListAsync(
partitionId,
0,
action.RequestTimeout,
cancellationToken),
helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
var removeUTRequestList = new List <Tuple <string, string> >();
Dictionary <Tuple <string, string>, Task> removeUTTaskDictionary = new Dictionary <Tuple <string, string>, Task>();
try
{
var stableReplicas = replicasResult.Where(r => r.ReplicaStatus == ServiceReplicaStatus.Ready).ToArray();
var stableReplicasToRemove = new List <StatefulServiceReplica>();
long replicasToRestartWithoutPrimary =
action.QuorumLossMode == QuorumLossMode.AllReplicas
? stableReplicas.Length - 1
: FabricCluster.GetWriteQuorumSize(replicasResult.Count);
foreach (var replica in stableReplicas)
{
StatefulServiceReplica statefulReplica = replica as StatefulServiceReplica;
ReleaseAssert.AssertIf(statefulReplica == null, "Service Replica is not of stateful type even though service is stateful");
if (statefulReplica.ReplicaRole != ReplicaRole.Primary)
{
replicasToRestartWithoutPrimary--;
}
if (replicasToRestartWithoutPrimary >= 0 || statefulReplica.ReplicaRole == ReplicaRole.Primary)
{
stableReplicasToRemove.Add(statefulReplica);
}
}
// for selected replicas, block reopen so that when we restart the replica (NOT remove the replica) it doesn't come up
var utTaskList = new List <Task>();
foreach (var statefulReplica in stableReplicasToRemove)
{
string nodeName = statefulReplica.NodeName;
UnreliableTransportBehavior behavior = new UnreliableTransportBehavior("*", "StatefulServiceReopen");
behavior.AddFilterForPartitionId(partitionId);
string behaviorName = "BlockStatefulServiceReopen_" + nodeName;
removeUTRequestList.Add(new Tuple <string, string>(nodeName, behaviorName));
utTaskList.Add(
FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() =>
testContext.FabricClient.TestManager.AddUnreliableTransportBehaviorAsync(
nodeName,
behaviorName,
behavior,
action.RequestTimeout,
cancellationToken),
helper.GetRemainingTime(),
cancellationToken));
}
await Task.WhenAll(utTaskList).ConfigureAwait(false);
// TODO: Wait for some time so that the unreliable transport behavior can be read from the files.
// Bug#2271465 - Unreliable transport through API should return only once the behavior has been successfully applied
await Task.Delay(TimeSpan.FromSeconds(5.0), cancellationToken);
var restartReplicaTaskList = new List <Task>();
foreach (var statefulReplica in stableReplicasToRemove)
{
ReplicaSelector replicaSelector = ReplicaSelector.ReplicaIdOf(PartitionSelector.PartitionIdOf(action.PartitionSelector.ServiceName, partitionId), statefulReplica.Id);
var restartReplicaAction = new RestartReplicaAction(replicaSelector)
{
CompletionMode = CompletionMode.DoNotVerify,
RequestTimeout = action.RequestTimeout,
ActionTimeout = helper.GetRemainingTime()
};
restartReplicaTaskList.Add(testContext.ActionExecutor.RunAsync(restartReplicaAction, cancellationToken));
}
await Task.WhenAll(restartReplicaTaskList).ConfigureAwait(false);
await AsyncWaiter.WaitAsync(action.QuorumLossDuration, cancellationToken).ConfigureAwait(false);
// validate
ServicePartitionList partitionsResult = await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.QueryManager.GetPartitionListAsync(
action.PartitionSelector.ServiceName,
null,
action.RequestTimeout,
cancellationToken),
FabricClientRetryErrors.GetPartitionListFabricErrors.Value,
helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
foreach (StatefulServicePartition partition in partitionsResult)
{
if (partition.PartitionInformation.Id == partitionId)
{
ReleaseAssert.AssertIf(partition.PartitionStatus != ServicePartitionStatus.InQuorumLoss, "Partition failed to be in Quorum Loss.");
break;
}
}
foreach (var removeUTParams in removeUTRequestList)
{
var currentParams = removeUTParams;
Task task = FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.TestManager.RemoveUnreliableTransportBehaviorAsync(
currentParams.Item1, /*nodeName*/
currentParams.Item2, /*behaviorName*/
action.RequestTimeout,
cancellationToken),
FabricClientRetryErrors.RemoveUnreliableTransportBehaviorErrors.Value,
helper.GetRemainingTime(),
cancellationToken);
removeUTTaskDictionary[currentParams] = task;
}
await Task.WhenAll(removeUTTaskDictionary.Values).ConfigureAwait(false);
// TODO: Wait for some time so that the removal of this unreliable transport behavior can be read from the files.
// Bug#2271465 - Unreliable transport through API should return only once the behavior has been successully applied
await Task.Delay(TimeSpan.FromSeconds(5.0), cancellationToken);
}
finally
{
var removeUTTaskList = new List <Task>();
foreach (var removeUTRequest in removeUTTaskDictionary)
{
var currentRemoveUTRequest = removeUTRequest;
if (currentRemoveUTRequest.Value == null || currentRemoveUTRequest.Value.IsFaulted)
{
removeUTTaskList.Add(
FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.TestManager.RemoveUnreliableTransportBehaviorAsync(
currentRemoveUTRequest.Key.Item1, /*nodeName*/
currentRemoveUTRequest.Key.Item2, /*behaviorName*/
action.RequestTimeout,
cancellationToken),
FabricClientRetryErrors.RemoveUnreliableTransportBehaviorErrors.Value,
helper.GetRemainingTime(),
cancellationToken));
}
}
Task.WhenAll(removeUTTaskList).Wait(cancellationToken);
// TODO: Wait for some time so that the removal of this unreliable transport behavior can be read from the files.
// Bug#2271465 - Unreliable transport through API should return only once the behavior has been successully applied
Task.Delay(TimeSpan.FromSeconds(5.0), cancellationToken).GetAwaiter().GetResult();
}
action.Result = new InvokeQuorumLossResult(getPartitionStateAction.Result);
this.ResultTraceString = StringHelper.Format("InvokeQuorumLossAction succeeded for {0} with QuorumLossMode = {1}", partitionId, action.QuorumLossMode);
}
public async Task <ValidationReport> EnsureStabilityWithReportAsync(TimeSpan maximumStabilizationTimeout, TimeSpan retryWait, CancellationToken ct)
{
TestabilityTrace.TraceSource.WriteInfo(TraceSource, "Ensuring that '{0}' is online with timeout '{1}'.", this.serviceName, maximumStabilizationTimeout);
bool checkQuorumLoss = (this.checkFlags & ValidationCheckFlag.CheckQuorumLoss) != 0;
// Load basic information about this service.
TestabilityTrace.TraceSource.WriteNoise(TraceSource, "Querying basic information for {0}.", this.serviceName);
await this.LoadPartitionAndReplicaCountAsync(ct);
DateTime startTime = DateTime.Now;
TimeoutHelper timer = new TimeoutHelper(maximumStabilizationTimeout);
bool success = false;
List <Guid> partitionsInQuorumLoss = new List <Guid>();
StringBuilder errorString = new StringBuilder();
int retryCount = 1;
while (!success && timer.GetRemainingTime() > TimeSpan.Zero)
{
TestabilityTrace.TraceSource.WriteInfo(TraceSource, "EnsureStabilityWithReportAsync(): retryCount='{0}', timer.GetRemainingTime()='{1}'", retryCount, timer.GetRemainingTime());
var nodes = await this.TestContext.FabricCluster.GetLatestNodeInfoAsync(this.requestTimeout, this.operationTimeout, ct);
// Empty error string and list of partitions in quorum loss
partitionsInQuorumLoss.Clear();
errorString.Clear();
success = true;
int totalPartitionsFound = 0;
bool stateful;
ReleaseAssert.AssertIfNot(this.isStateful.TryGetValue(out stateful), "isStateful flag is not available");
bool checkTarget = (this.checkFlags & ValidationCheckFlag.CheckTargetReplicaSetSize) != 0;
bool checkInBuild = (this.checkFlags & ValidationCheckFlag.CheckInBuildReplica) != 0;
if (stateful)
{
var partitionDictionary = await this.QueryPartitionAndReplicaResultAsyncStateful(ct);
totalPartitionsFound = partitionDictionary.Count();
foreach (KeyValuePair <Partition, StatefulServiceReplica[]> partition in partitionDictionary)
{
bool partitionIsReady = partition.Key.PartitionStatus == ServicePartitionStatus.Ready;
if (!partitionIsReady)
{
var message = StringHelper.Format("Partition '{0}' is not Ready", partition.Key.PartitionId());
TestabilityTrace.TraceSource.WriteInfo(TraceSource, "{0}", message);
errorString.AppendLine(message);
}
if (partition.Key.PartitionStatus != ServicePartitionStatus.InQuorumLoss)
{
int validCount = 0;
int inBuildReplicas = 0;
foreach (StatefulServiceReplica replica in partition.Value)
{
if (replica.ReplicaStatus == ServiceReplicaStatus.Ready &&
(replica.ReplicaRole == ReplicaRole.Primary || replica.ReplicaRole == ReplicaRole.ActiveSecondary))
{
++validCount;
}
if (replica.ReplicaStatus == ServiceReplicaStatus.InBuild)
{
++inBuildReplicas;
var message = StringHelper.Format("Replica {0} for partition '{1}' is InBuild", replica.Id, partition.Key.PartitionId());
TestabilityTrace.TraceSource.WriteInfo(TraceSource, "{0}", message);
errorString.AppendLine(message);
}
}
bool targetAchieved = this.CheckReplicaSetSize(partition.Key.PartitionInformation.Id, validCount, startTime, nodes, errorString);
if (!partitionIsReady ||
(checkInBuild && inBuildReplicas > 0) ||
(checkTarget && !targetAchieved))
{
success = false;
}
}
else
{
partitionsInQuorumLoss.Add(partition.Key.PartitionInformation.Id);
}
}
}
else
{
int targetInstanceCount = 0;
ReleaseAssert.AssertIf(!this.targetReplicaSetSize.TryGetValue(out targetInstanceCount), "targetReplicaSetSize for service: {0} should have been populated at this point.", this.serviceName);
bool placementConstraintsDefined = false;
try
{
// Get the service description to find out if there are placement constraints on the service
ServiceDescription result = await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => this.TestContext.FabricClient.ServiceManager.GetServiceDescriptionAsync(
this.serviceName,
this.requestTimeout,
ct),
this.operationTimeout,
ct).ConfigureAwait(false);
ThrowIf.IsTrue(result == null, "A description must be associated with the service: {0}", this.serviceName);
placementConstraintsDefined = !string.IsNullOrEmpty(result.PlacementConstraints);
}
catch (UnauthorizedAccessException)
{
ServiceGroupDescription groupDescription = await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => this.TestContext.FabricClient.ServiceGroupManager.GetServiceGroupDescriptionAsync(
this.serviceName,
this.requestTimeout,
ct),
this.operationTimeout,
ct).ConfigureAwait(false);
ThrowIf.IsTrue(groupDescription == null, "A description must be associated with the service group: {0}", this.serviceName);
placementConstraintsDefined = !string.IsNullOrEmpty(groupDescription.ServiceDescription.PlacementConstraints);
}
// If a stateless service has instance count == -1 and it has placement constraints such
// that the possible number of instances cannot match the total number of nodes,
// we need to find out the number of eligible nodes for the service which is tracked by RDBug 8993319.
// Until RDBug 8993319 is fixed, we take the presence of placement constraints into consideration to make the
// validation more accurate.
if (targetInstanceCount == -1 && placementConstraintsDefined)
{
checkTarget = false;
}
var partitionDictionary = await this.QueryPartitionAndReplicaResultAsyncStateless(timer.GetRemainingTime(), ct);
totalPartitionsFound = partitionDictionary.Count();
foreach (KeyValuePair <Partition, StatelessServiceInstance[]> partition in partitionDictionary)
{
bool partitionIsReady = partition.Key.PartitionStatus == ServicePartitionStatus.Ready;
if (!partitionIsReady)
{
var message = StringHelper.Format("Partition '{0}' is not Ready", partition.Key.PartitionId());
TestabilityTrace.TraceSource.WriteInfo(TraceSource, "{0}", message);
errorString.AppendLine(message);
}
int validCount = 0;
foreach (StatelessServiceInstance instance in partition.Value)
{
if (instance.ReplicaStatus == ServiceReplicaStatus.Ready)
{
++validCount;
}
}
bool targetAchieved = this.CheckReplicaSetSize(partition.Key.PartitionInformation.Id, validCount, startTime, nodes, errorString);
if (!partitionIsReady ||
(checkTarget && !targetAchieved))
{
success = false;
}
}
}
if (!this.ValidatePartitionCount(totalPartitionsFound))
{
success = false;
}
if (partitionsInQuorumLoss.Count > 0 && checkQuorumLoss)
{
string paritionIds = string.Join(",", partitionsInQuorumLoss.ToArray());
var message = StringHelper.Format("Partitions '{0}' in quorum loss for service {1}", paritionIds, this.serviceName);
TestabilityTrace.TraceSource.WriteInfo(TraceSource, "{0}", message);
errorString.AppendLine(message);
success = false;
}
if (!success)
{
if (retryCount % 10 == 0)
{
TestabilityTrace.TraceSource.WriteWarning(TraceSource, "Service {0} validation failed due to issues below, will retry: \n{1}", this.serviceName, errorString);
}
// Delay before querying again so we allow some time for state to change - don't spam the node
await AsyncWaiter.WaitAsync(retryWait, ct).ConfigureAwait(false);
}
retryCount++;
}
if (partitionsInQuorumLoss.Count > 0)
{
string partitionIds = string.Join(",", partitionsInQuorumLoss.ToArray());
TestabilityTrace.TraceSource.WriteInfo(TraceSource, "Partitions in quorum loss for service {0} are '{1}'", this.serviceName, partitionIds);
if (checkQuorumLoss)
{
throw new FabricValidationException(StringHelper.Format(StringResources.Error_PartitionsInQuorumLoss, partitionIds, this.serviceName));
}
}
if (!success)
{
return(new ValidationReport(
true,
StringHelper.Format(StringResources.Error_ServiceNotStable, this.serviceName, maximumStabilizationTimeout, errorString)));
}
else
{
return(ValidationReport.Default);
}
}
public async Task <ValidationReport> ValidateHealthWithReportAsync(TimeSpan maximumStabilizationTimeout, TimeSpan retryWait, CancellationToken ct)
{
TestabilityTrace.TraceSource.WriteInfo(TraceSource, "Validating that '{0}' is healthy with timeout '{1}'.", this.serviceName, maximumStabilizationTimeout);
TimeoutHelper timer = new TimeoutHelper(maximumStabilizationTimeout);
bool success = false;
string healthinfo = string.Empty;
int retryCount = 1;
while (!success && timer.GetRemainingTime() > TimeSpan.Zero)
{
TestabilityTrace.TraceSource.WriteInfo(TraceSource, "ValidateHealthWithReportAsync(): retryCount='{0}', timer.GetRemainingTime()='{1}'", retryCount, timer.GetRemainingTime());
healthinfo = string.Empty;
if (this.TestContext == null)
{
Console.WriteLine("testcontext is null");
}
ReleaseAssert.AssertIfNull(this.TestContext, "test context");
ReleaseAssert.AssertIfNull(this.serviceName, "serviceName");
ReleaseAssert.AssertIfNull(FabricClientRetryErrors.GetEntityHealthFabricErrors.Value, "health error code");
ApplicationHealthPolicy healthPolicy = new ApplicationHealthPolicy();
var serviceHealthResult = await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() =>
this.TestContext.FabricClient.HealthManager.GetServiceHealthAsync(
this.serviceName,
healthPolicy,
this.requestTimeout,
ct),
FabricClientRetryErrors.GetEntityHealthFabricErrors.Value,
timer.GetRemainingTime(),
ct).ConfigureAwait(false);
bool checkError = (this.checkFlags & ValidationCheckFlag.CheckError) != 0;
bool checkWarning = (this.checkFlags & ValidationCheckFlag.CheckWarning) != 0;
if ((checkError && serviceHealthResult.AggregatedHealthState == HealthState.Error) ||
(checkWarning && serviceHealthResult.AggregatedHealthState == HealthState.Warning) ||
serviceHealthResult.AggregatedHealthState == HealthState.Invalid ||
serviceHealthResult.AggregatedHealthState == HealthState.Unknown)
{
TestabilityTrace.TraceSource.WriteInfo(TraceSource, "{0} is health state is {1}. Will Retry check", this.serviceName, serviceHealthResult.AggregatedHealthState);
healthinfo = await this.GetUnhealthyItemsAsync(serviceHealthResult, timer, ct).ConfigureAwait(false);
TestabilityTrace.TraceSource.WriteInfo(TraceSource, healthinfo);
}
else
{
success = true;
}
if (!success)
{
if (retryCount % 10 == 0)
{
TestabilityTrace.TraceSource.WriteWarning(TraceSource, "Service {0} health validation failed due to issues below, will retry: \n{1}", this.serviceName, healthinfo);
}
// Delay before querying again so we allow some time for state to change - don't spam the node
await AsyncWaiter.WaitAsync(retryWait);
}
retryCount++;
}
if (!success)
{
return(new ValidationReport(true, StringHelper.Format(StringResources.Error_ServiceNotHealthy, serviceName, maximumStabilizationTimeout, healthinfo)));
}
else
{
return(ValidationReport.Default);
}
}
protected override async Task ExecuteActionAsync(FabricTestContext testContext, InvokeDataLossAction action, CancellationToken cancellationToken)
{
ThrowIf.Null(action.PartitionSelector, "PartitionSelector");
var helper = new TimeoutHelper(action.ActionTimeout);
ServiceDescription result = await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.ServiceManager.GetServiceDescriptionAsync(
action.PartitionSelector.ServiceName,
action.RequestTimeout,
cancellationToken),
helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
if (result.Kind != ServiceDescriptionKind.Stateful)
{
throw new InvalidOperationException(StringHelper.Format(StringResources.Error_InvalidServiceTypeTestability, "DataLoss", "Stateful", action.PartitionSelector.ServiceName, "Stateless"));
}
var getPartitionStateAction = new GetSelectedPartitionStateAction(action.PartitionSelector)
{
RequestTimeout = action.RequestTimeout,
ActionTimeout = helper.GetRemainingTime()
};
await testContext.ActionExecutor.RunAsync(getPartitionStateAction, cancellationToken).ConfigureAwait(false);
Guid partitionId = getPartitionStateAction.Result.PartitionId;
long preDataLossNumber = 0;
ServicePartitionList partitionsResult = await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.QueryManager.GetPartitionListAsync(
action.PartitionSelector.ServiceName,
null,
action.RequestTimeout,
cancellationToken),
helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
bool partitionFound = false;
foreach (StatefulServicePartition partition in partitionsResult)
{
if (partition.PartitionInformation.Id == partitionId)
{
preDataLossNumber = partition.PrimaryEpoch.DataLossNumber;
partitionFound = true;
break;
}
}
if (!partitionFound)
{
throw new FabricException(StringHelper.Format(StringResources.Error_PartitionNotFound), FabricErrorCode.PartitionNotFound);
}
long postDataLossNumber = preDataLossNumber;
do
{
ActionTraceSource.WriteInfo(
TraceType,
"InvokeDataLossAction action pending time:{0}",
helper.GetRemainingTime());
if (helper.GetRemainingTime() <= TimeSpan.Zero)
{
throw new TimeoutException(StringHelper.Format(StringResources.Error_TestabilityActionTimeout, "InvokeDataLoss", partitionId));
}
ServiceReplicaList replicasResult = await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.QueryManager.GetReplicaListAsync(
partitionId,
0,
action.RequestTimeout,
cancellationToken),
helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
ServiceReplicaList fmReplicasResult = await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.QueryManager.GetReplicaListAsync(
Constants.FmPartitionId,
0,
action.RequestTimeout,
cancellationToken),
helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
string fmPrimaryNodeName = string.Empty;
var readyFMReplicas = fmReplicasResult.Where(r => r.ReplicaStatus == ServiceReplicaStatus.Ready).ToArray();
foreach (var replica in readyFMReplicas)
{
StatefulServiceReplica statefulReplica = replica as StatefulServiceReplica;
ReleaseAssert.AssertIf(statefulReplica == null, "FM Replica is not a stateful replica");
if (statefulReplica.ReplicaRole == ReplicaRole.Primary)
{
fmPrimaryNodeName = replica.NodeName;
}
}
if (string.IsNullOrEmpty(fmPrimaryNodeName))
{
throw new FabricException(StringHelper.Format(StringResources.Error_PartitionPrimaryNotReady, "FailoverManager"), FabricErrorCode.NotReady);
}
UnreliableTransportBehavior behavior = new UnreliableTransportBehavior("*", "DoReconfiguration");
behavior.AddFilterForPartitionId(partitionId);
string behaviorName = "BlockDoReconfiguration";
await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.TestManager.AddUnreliableTransportBehaviorAsync(
fmPrimaryNodeName,
behaviorName,
behavior,
action.RequestTimeout,
cancellationToken),
helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
// TODO: Wait for some time so that the unreliable transport behavior can be read from the files.
// Bug#2271465 - Unreliable transport through API should return only once the behavior has been successfully applied
await Task.Delay(TimeSpan.FromSeconds(5.0), cancellationToken).ConfigureAwait(false);
bool triedToRemovedBehavior = false;
try
{
var stableReplicasToRemove = replicasResult.Where(r => r.ReplicaStatus == ServiceReplicaStatus.Ready).ToArray();
ActionTraceSource.WriteInfo(TraceType, "Total number of replicas found {0}:{1}", replicasResult.Count(), stableReplicasToRemove.Count());
int replicasToRestartWithoutPrimary =
action.DataLossMode == DataLossMode.FullDataLoss
? stableReplicasToRemove.Length - 1
: (stableReplicasToRemove.Length + 1) / 2 - 1;
foreach (var replica in stableReplicasToRemove)
{
var currentReplica = replica;
StatefulServiceReplica statefulReplica = currentReplica as StatefulServiceReplica;
ReleaseAssert.AssertIf(statefulReplica == null, "Service Replica is not of stateful type even though service is stateful");
ActionTraceSource.WriteInfo(
TraceType,
"Inspecting replica {0}:{1} with role {2} and status {3} to induce data loss",
currentReplica.Id,
partitionId,
statefulReplica.ReplicaRole,
statefulReplica.ReplicaStatus);
if (statefulReplica.ReplicaRole != ReplicaRole.Primary)
{
replicasToRestartWithoutPrimary--;
}
if (replicasToRestartWithoutPrimary >= 0 || statefulReplica.ReplicaRole == ReplicaRole.Primary)
{
ActionTraceSource.WriteInfo(TraceType, "Removing replica {0}:{1} to induce data loss", currentReplica.Id, partitionId);
await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.FaultManager.RemoveReplicaAsync(
currentReplica.NodeName,
partitionId,
currentReplica.Id,
CompletionMode.DoNotVerify,
false, /*force remove*/
action.RequestTimeout.TotalSeconds,
cancellationToken),
helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
}
}
triedToRemovedBehavior = true;
await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.TestManager.RemoveUnreliableTransportBehaviorAsync(
fmPrimaryNodeName,
behaviorName,
action.RequestTimeout,
cancellationToken),
FabricClientRetryErrors.RemoveUnreliableTransportBehaviorErrors.Value,
helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
// TODO: Wait for some time so that the removal of this unreliable transport behavior can be read from the files.
// Bug#2271465 - Unreliable transport through API should return only once the behavior has been successully applied
await Task.Delay(TimeSpan.FromSeconds(5.0), cancellationToken).ConfigureAwait(false);
// retry check for whether data loss number has increased 5 times else do the entire process again
const int maxRetryCount = 5;
int retryCount = 0;
do
{
partitionsResult = await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.QueryManager.GetPartitionListAsync(
action.PartitionSelector.ServiceName,
null,
action.RequestTimeout,
cancellationToken),
FabricClientRetryErrors.GetPartitionListFabricErrors.Value,
helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
partitionFound = false;
foreach (StatefulServicePartition partition in partitionsResult)
{
if (partition.PartitionInformation.Id == partitionId)
{
postDataLossNumber = partition.PrimaryEpoch.DataLossNumber;
partitionFound = true;
break;
}
}
if (!partitionFound)
{
throw new FabricException(StringHelper.Format(StringResources.Error_PartitionNotFound), FabricErrorCode.PartitionNotFound);
}
ActionTraceSource.WriteInfo(
TraceType,
"Checking data loss numbers for partition {0} with retryCount {1}. Current numbers {2}:{3}",
partitionId,
retryCount,
preDataLossNumber,
postDataLossNumber);
if (postDataLossNumber != preDataLossNumber)
{
break;
}
await AsyncWaiter.WaitAsync(TimeSpan.FromSeconds(5), cancellationToken);
++retryCount;
} while (retryCount < maxRetryCount);
}
finally
{
if (!triedToRemovedBehavior)
{
ActionTraceSource.WriteWarning(TraceType, "Exception after adding behavior to block messages. Removing behavior synchronously");
FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.TestManager.RemoveUnreliableTransportBehaviorAsync(
fmPrimaryNodeName,
behaviorName,
action.RequestTimeout,
cancellationToken),
FabricClientRetryErrors.RemoveUnreliableTransportBehaviorErrors.Value,
helper.GetRemainingTime(),
cancellationToken).GetAwaiter().GetResult();
// TODO: Wait for some time so that the removal of this unreliable transport behavior can be read from the files.
// Bug#2271465 - Unreliable transport through API should return only once the behavior has been successully applied
Task.Delay(TimeSpan.FromSeconds(5.0), cancellationToken).GetAwaiter().GetResult();
}
}
}while (postDataLossNumber == preDataLossNumber);
ActionTraceSource.WriteInfo(
TraceType,
"InvokeDataLossAction action completed postDataLossNumber:{0}, preDataLossNumber:{1}",
postDataLossNumber, preDataLossNumber);
action.Result = new InvokeDataLossResult(getPartitionStateAction.Result);
this.ResultTraceString = StringHelper.Format("InvokeDataLossAction succeeded for {0} with DatalossMode = {1}", partitionId, action.DataLossMode);
}
protected override async Task ExecuteActionAsync(FabricTestContext testContext, RemoveReplicaAction action, CancellationToken cancellationToken)
{
TimeoutHelper helper = new TimeoutHelper(action.ActionTimeout);
string nodeName = action.NodeName;
Guid? partitionId = action.PartitionId;
long? replicaId = action.ReplicaId;
SelectedReplica replicaSelectorResult = SelectedReplica.None;
if (string.IsNullOrEmpty(nodeName) ||
!partitionId.HasValue ||
!replicaId.HasValue)
{
ThrowIf.Null(action.ReplicaSelector, "ReplicaSelector");
var getReplicaStateAction = new GetSelectedReplicaStateAction(action.ReplicaSelector)
{
RequestTimeout = action.RequestTimeout,
ActionTimeout = helper.GetRemainingTime()
};
await testContext.ActionExecutor.RunAsync(getReplicaStateAction, cancellationToken).ConfigureAwait(false);
var replicaStateActionResult = getReplicaStateAction.Result;
ReleaseAssert.AssertIf(replicaStateActionResult == null, "replicaStateActionResult cannot be null");
replicaSelectorResult = replicaStateActionResult.Item1;
partitionId = replicaStateActionResult.Item1.SelectedPartition.PartitionId;
Replica replicaStateResult = replicaStateActionResult.Item2;
ReleaseAssert.AssertIf(replicaStateResult == null, "replicaStateResult cannot be null");
nodeName = replicaStateResult.NodeName;
replicaId = replicaStateResult.Id;
}
ThrowIf.IsFalse(partitionId.HasValue, "PartitionID");
ThrowIf.IsFalse(replicaId.HasValue, "ReplicaID");
bool forceRemove = action.ForceRemove;
await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.ServiceManager.RemoveReplicaAsync(
nodeName,
partitionId.Value,
replicaId.Value,
forceRemove,
action.RequestTimeout,
cancellationToken),
FabricClientRetryErrors.RemoveReplicaErrors.Value,
helper.GetRemainingTime(),
cancellationToken);
if (action.CompletionMode == CompletionMode.Verify)
{
// Check that replica on selected node has been removed i.e. the replica id does not exist anymore.
bool success = false;
while (helper.GetRemainingTime() > TimeSpan.Zero)
{
var replicasResult = await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.QueryManager.GetReplicaListAsync(
partitionId.Value,
replicaId.Value,
action.RequestTimeout,
cancellationToken),
helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
bool dropped = replicasResult.Count == 0;
if (!dropped)
{
// Since we added a replica filter the result should contain the replica or none
ReleaseAssert.AssertIf(replicasResult.Count > 1, "More than 1 replica returned with replica filter {0}:{1}", partitionId.Value, replicaId.Value);
ReleaseAssert.AssertIf(replicasResult[0].Id != replicaId, "Incorrect replica Id {0} returned by query instead of {1}", replicasResult[0].Id, replicaId);
dropped = replicasResult[0].ReplicaStatus == ServiceReplicaStatus.Dropped;
}
if (dropped)
{
success = true;
break;
}
ActionTraceSource.WriteInfo(TraceSource, "Replica = {0}:{1} not yet completely removed. Retrying...", partitionId.Value, replicaId.Value);
await AsyncWaiter.WaitAsync(TimeSpan.FromSeconds(5), cancellationToken);
}
if (!success)
{
throw new TimeoutException(StringHelper.Format(StringResources.Error_TestabilityActionTimeout,
"RemoveReplica",
StringHelper.Format("{0}:{1}", partitionId.Value, replicaId.Value)));
}
}
action.Result = new RemoveReplicaResult(replicaSelectorResult);
ResultTraceString = StringHelper.Format(
"RemoveReplicaOrInstance succeeded by removing replica {0}:{1} on node {2} with CompletionMode {3}",
partitionId.Value,
replicaId.Value,
nodeName,
action.CompletionMode);
}
protected override async Task ExecuteActionAsync(FabricTestContext testContext, RestartDeployedCodePackageAction action, CancellationToken cancellationToken)
{
this.helper = new TimeoutHelper(action.ActionTimeout);
string nodeName = action.NodeName;
Uri applicationName = action.ApplicationName;
string serviceManifestName = action.ServiceManifestName;
string servicePackageActivationId = action.ServicePackageActivationId;
string codePackageName = action.CodePackageName;
SelectedReplica replicaSelectorResult = SelectedReplica.None;
ThrowIf.Null(applicationName, "ApplicationName");
if (string.IsNullOrEmpty(nodeName) ||
string.IsNullOrEmpty(serviceManifestName) ||
string.IsNullOrEmpty(codePackageName))
{
ThrowIf.Null(action.ReplicaSelector, "ReplicaSelector");
var getReplicaStateAction = new GetSelectedReplicaStateAction(action.ReplicaSelector)
{
RequestTimeout = action.RequestTimeout,
ActionTimeout = this.helper.GetRemainingTime()
};
await testContext.ActionExecutor.RunAsync(getReplicaStateAction, cancellationToken).ConfigureAwait(false);
var replicaStateActionResult = getReplicaStateAction.Result;
ReleaseAssert.AssertIf(replicaStateActionResult == null, "replicaStateActionResult cannot be null");
replicaSelectorResult = replicaStateActionResult.Item1;
ReleaseAssert.AssertIf(replicaSelectorResult == null || replicaSelectorResult.SelectedPartition == null,
"replicaSelectorResult cannot be null or for a non-null replicaSelectorResult, the selected partition must be non-null");
Guid partitionId = replicaStateActionResult.Item1.SelectedPartition.PartitionId;
Replica replicaStateResult = replicaStateActionResult.Item2;
ReleaseAssert.AssertIf(replicaStateResult == null, "replicaStateResult cannot be null");
nodeName = replicaStateResult.NodeName;
var deployedReplicaListResult = await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync <DeployedServiceReplicaList>(
() => testContext.FabricClient.QueryManager.GetDeployedReplicaListAsync(
nodeName,
applicationName,
null,
partitionId,
action.RequestTimeout,
cancellationToken),
this.helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
DeployedServiceReplica selectedReplica = deployedReplicaListResult.FirstOrDefault(r => r.Partitionid == partitionId);
if (selectedReplica == null)
{
throw new FabricException(
StringHelper.Format(StringResources.Error_DidNotFindDeployedReplicaOnNode, partitionId, nodeName),
FabricErrorCode.ReplicaDoesNotExist);
}
serviceManifestName = selectedReplica.ServiceManifestName;
servicePackageActivationId = selectedReplica.ServicePackageActivationId;
codePackageName = selectedReplica.CodePackageName;
}
ActionTraceSource.WriteInfo(TraceSource, "SelectedReplica: serviceManifestName: {0}, servicePackageActivationId: {1}, codePackageName: {2}", serviceManifestName, servicePackageActivationId, codePackageName);
DeployedCodePackage deployedCodePackageListResult = await this.GetCodePackageInfoAsync(testContext, nodeName, applicationName, serviceManifestName, servicePackageActivationId, codePackageName, action, cancellationToken).ConfigureAwait(false);
var codepackageEntrypointToRestart = GetCodepackageEntrypointToRestart(action, deployedCodePackageListResult);
await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.FaultManager.RestartDeployedCodePackageUsingNodeNameAsync(
nodeName,
applicationName,
serviceManifestName,
servicePackageActivationId,
codePackageName,
codepackageEntrypointToRestart.EntryPoint.CodePackageInstanceId,
action.RequestTimeout,
cancellationToken),
this.helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
if (action.CompletionMode == CompletionMode.Verify)
{
bool success = false;
while (this.helper.GetRemainingTime() > TimeSpan.Zero)
{
var deployedCodePackageListResultAfterRestart = await this.GetCodePackageInfoAsync(testContext, nodeName, applicationName, serviceManifestName, servicePackageActivationId, codePackageName, action, cancellationToken).ConfigureAwait(false);
if (deployedCodePackageListResultAfterRestart != null)
{
var entryPointAfterRestart = codepackageEntrypointToRestart.EntryPointType == EntryPointType.Main ? deployedCodePackageListResultAfterRestart.EntryPoint : deployedCodePackageListResultAfterRestart.SetupEntryPoint;
if (entryPointAfterRestart != null && entryPointAfterRestart.CodePackageInstanceId > codepackageEntrypointToRestart.EntryPoint.CodePackageInstanceId && entryPointAfterRestart.EntryPointStatus == EntryPointStatus.Started)
{
success = true;
break;
}
}
ActionTraceSource.WriteInfo(TraceSource, "CodePackage = {0}:{1}:{2} not yet restarted. Retrying...", nodeName, applicationName, codePackageName);
await AsyncWaiter.WaitAsync(TimeSpan.FromSeconds(5), cancellationToken).ConfigureAwait(false);
}
if (!success)
{
throw new TimeoutException(StringHelper.Format(StringResources.Error_TestabilityActionTimeout,
"RestartDeployedCodePackage",
applicationName));
}
}
action.Result = new RestartDeployedCodePackageResult(
nodeName,
applicationName,
serviceManifestName,
servicePackageActivationId,
codePackageName,
codepackageEntrypointToRestart.EntryPoint.CodePackageInstanceId,
replicaSelectorResult);
ResultTraceString = StringHelper.Format("RestartCodePackageAction succeeded for {0}:{1}:{2} with CompletionMode = {3}", nodeName, applicationName, codePackageName, action.CompletionMode);
}
protected override async Task ExecuteActionAsync(FabricTestContext testContext, RestartNodeAction action, CancellationToken cancellationToken)
{
ActionTraceSource.WriteInfo(TraceSource, "Enter RestartNodeAction/ExecuteActionAsync: operationTimeout='{0}', requestTimeout='{1}'", action.ActionTimeout, action.RequestTimeout);
this.helper = new TimeoutHelper(action.ActionTimeout);
SelectedReplica selectedReplica = SelectedReplica.None;
string nodeName = action.NodeName;
BigInteger nodeInstance = action.NodeInstance;
bool createFabricDump = action.CreateFabricDump;
if (string.IsNullOrEmpty(nodeName))
{
ThrowIf.Null(action.ReplicaSelector, "ReplicaSelector");
var getReplicaStateAction = new GetSelectedReplicaStateAction(action.ReplicaSelector)
{
RequestTimeout = action.RequestTimeout,
ActionTimeout = helper.GetRemainingTime()
};
await testContext.ActionExecutor.RunAsync(getReplicaStateAction, cancellationToken).ConfigureAwait(false);
var replicaStateActionResult = getReplicaStateAction.Result;
ReleaseAssert.AssertIf(replicaStateActionResult == null, "replicaStateActionResult cannot be null");
selectedReplica = replicaStateActionResult.Item1;
Replica replicaStateResult = replicaStateActionResult.Item2;
ReleaseAssert.AssertIf(replicaStateResult == null, "replicaStateResult cannot be null");
nodeName = replicaStateResult.NodeName;
nodeInstance = BigInteger.MinusOne;
}
if (nodeInstance == BigInteger.MinusOne)
{
var nodeInfo = await this.GetCurrentNodeInfoAsync(testContext, nodeName, action, cancellationToken).ConfigureAwait(false);
nodeInstance = nodeInfo.NodeInstanceId;
}
await FabricClientRetryHelper.ExecuteFabricActionWithRetryAsync(
() => testContext.FabricClient.FaultManager.RestartNodeUsingNodeNameAsync(
nodeName,
nodeInstance,
createFabricDump,
action.RequestTimeout,
cancellationToken),
this.helper.GetRemainingTime(),
cancellationToken).ConfigureAwait(false);
if (action.CompletionMode == CompletionMode.Verify)
{
bool success = false;
while (this.helper.GetRemainingTime() > TimeSpan.Zero)
{
var nodeInfo = await this.GetCurrentNodeInfoAsync(testContext, nodeName, action, cancellationToken).ConfigureAwait(false);
if (nodeInfo.NodeInstanceId > nodeInstance && nodeInfo.IsNodeUp)
{
success = true;
break;
}
ActionTraceSource.WriteInfo(TraceSource, "NodeName = {0} not yet restarted. '{1}' seconds remain. Retrying...", nodeName, this.helper.GetRemainingTime().TotalSeconds);
await AsyncWaiter.WaitAsync(TimeSpan.FromSeconds(5), cancellationToken);
}
if (!success)
{
throw new TimeoutException(StringHelper.Format(StringResources.Error_TestabilityActionTimeout,
"RestartNode",
nodeName));
}
}
// create result
action.Result = new RestartNodeResult(selectedReplica, new NodeResult(nodeName, nodeInstance));
ResultTraceString = StringHelper.Format("RestartNodeAction succeeded for {0}:{1} with CompletionMode = {2}", nodeName, nodeInstance, action.CompletionMode);
}