def do_view_change_with_propagate_primary_on_one_delayed_node(
slow_node, nodes, looper, sdk_pool_handle, sdk_wallet_client):
slow_stasher = slow_node.nodeIbStasher
fast_nodes = [n for n in nodes if n != slow_node]
stashers = [n.nodeIbStasher for n in nodes]
# Get last prepared certificate in pool
lpc = last_prepared_certificate(nodes)
# Get pool current view no
view_no = lpc[0]
with delay_rules(slow_stasher, icDelay()):
with delay_rules(slow_stasher, vcd_delay()):
with delay_rules(stashers, cDelay()):
# Send request
request = sdk_send_random_request(looper, sdk_pool_handle, sdk_wallet_client)
# Wait until this request is prepared on N-f nodes
looper.run(eventually(check_last_prepared_certificate_on_quorum, nodes, (lpc[0], lpc[1] + 1)))
# Trigger view change
for n in nodes:
n.view_changer.on_master_degradation()
# Wait until view change is completed on all nodes except slow one
waitForViewChange(looper,
fast_nodes,
expectedViewNo=view_no + 1,
customTimeout=waits.expectedPoolViewChangeStartedTimeout(len(nodes)))
wait_for_elections_done_on_given_nodes(looper,
fast_nodes,
getRequiredInstances(len(nodes)),
timeout=waits.expectedPoolElectionTimeout(len(nodes)))
# Now all the nodes receive Commits
# The slow node will accept Commits and order the 3PC-batch in the old view
looper.runFor(waits.expectedOrderingTime(getNoInstances(len(nodes))))
# Now slow node receives ViewChangeDones
waitForViewChange(looper,
[slow_node],
expectedViewNo=view_no + 1,
customTimeout=waits.expectedPoolViewChangeStartedTimeout(len(nodes)))
wait_for_elections_done_on_given_nodes(looper,
[slow_node],
getRequiredInstances(len(nodes)),
timeout=waits.expectedPoolElectionTimeout(len(nodes)))
# Now slow node receives InstanceChanges but discards them because already
# started propagate primary to the same view.
# Finish request gracefully
sdk_get_reply(looper, request)
def do_view_change_with_propagate_primary_on_one_delayed_node(
slow_node, nodes, looper, sdk_pool_handle, sdk_wallet_client):
slow_stasher = slow_node.nodeIbStasher
fast_nodes = [n for n in nodes if n != slow_node]
stashers = [n.nodeIbStasher for n in nodes]
# Get last prepared certificate in pool
lpc = last_prepared_certificate(nodes)
# Get pool current view no
view_no = lpc[0]
with delay_rules(slow_stasher, icDelay()):
with delay_rules(slow_stasher, vcd_delay()):
with delay_rules(stashers, cDelay()):
# Send request
request = sdk_send_random_request(looper, sdk_pool_handle, sdk_wallet_client)
# Wait until this request is prepared on N-f nodes
looper.run(eventually(check_last_prepared_certificate_on_quorum, nodes, (lpc[0], lpc[1] + 1)))
# Trigger view change
for n in nodes:
n.view_changer.on_master_degradation()
# Wait until view change is completed on all nodes except slow one
waitForViewChange(looper,
fast_nodes,
expectedViewNo=view_no + 1,
customTimeout=waits.expectedPoolViewChangeStartedTimeout(len(nodes)))
wait_for_elections_done_on_given_nodes(looper,
fast_nodes,
getRequiredInstances(len(nodes)),
timeout=waits.expectedPoolElectionTimeout(len(nodes)))
# Now all the nodes receive Commits
# The slow node will accept Commits and order the 3PC-batch in the old view
looper.runFor(waits.expectedOrderingTime(getNoInstances(len(nodes))))
# Now slow node receives ViewChangeDones
waitForViewChange(looper,
[slow_node],
expectedViewNo=view_no + 1,
customTimeout=waits.expectedPoolViewChangeStartedTimeout(len(nodes)))
wait_for_elections_done_on_given_nodes(looper,
[slow_node],
getRequiredInstances(len(nodes)),
timeout=waits.expectedPoolElectionTimeout(len(nodes)))
# Now slow node receives InstanceChanges but discards them because already
# started propagate primary to the same view.
# Finish request gracefully
sdk_get_reply(looper, request)
def test_quorum_after_f_plus_2_nodes_including_primary_turned_off_and_later_on(
looper, allPluginsPath, tdir, tconf,
txnPoolNodeSet, wallet1, client1):
nodes = txnPoolNodeSet
request1 = sendRandomRequest(wallet1, client1)
waitForSufficientRepliesForRequests(looper, client1, requests=[request1])
stop_node(nodes[0], looper, nodes)
waitForViewChange(looper, nodes[1:], expectedViewNo=1)
ensureElectionsDone(looper, nodes[1:],
numInstances=getRequiredInstances(nodeCount))
request2 = sendRandomRequest(wallet1, client1)
waitForSufficientRepliesForRequests(looper, client1, requests=[request2])
stop_node(nodes[1], looper, nodes)
looper.runFor(tconf.ToleratePrimaryDisconnection +
waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[2:], expectedViewNo=1)
request3 = sendRandomRequest(wallet1, client1)
verify_request_not_replied_and_not_ordered(request3, looper, client1, nodes)
stop_node(nodes[2], looper, nodes)
looper.runFor(tconf.ToleratePrimaryDisconnection +
waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[3:], expectedViewNo=1)
request4 = sendRandomRequest(wallet1, client1)
verify_request_not_replied_and_not_ordered(request4, looper, client1, nodes)
nodes[2] = start_stopped_node(nodes[2], looper, tconf, tdir, allPluginsPath)
looper.runFor(waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[3:], expectedViewNo=1)
request5 = sendRandomRequest(wallet1, client1)
verify_request_not_replied_and_not_ordered(request5, looper, client1, nodes)
nodes[1] = start_stopped_node(nodes[1], looper, tconf, tdir, allPluginsPath)
ensureElectionsDone(looper, nodes[1:],
numInstances=getRequiredInstances(nodeCount))
checkViewNoForNodes(nodes[1:], expectedViewNo=1)
request6 = sendRandomRequest(wallet1, client1)
waitForSufficientRepliesForRequests(looper, client1, requests=[request6])
nodes[0] = start_stopped_node(nodes[0], looper, tconf, tdir, allPluginsPath)
ensureElectionsDone(looper, nodes,
numInstances=getRequiredInstances(nodeCount))
checkViewNoForNodes(nodes, expectedViewNo=1)
request7 = sendRandomRequest(wallet1, client1)
waitForSufficientRepliesForRequests(looper, client1, requests=[request7])
def ensureElectionsDone(
looper: Looper,
nodes: Sequence[TestNode],
retryWait: float = None, # seconds
customTimeout: float = None,
numInstances: int = None) -> Sequence[TestNode]:
# TODO: Change the name to something like `ensure_primaries_selected`
# since there might not always be an election, there might be a round
# robin selection
"""
Wait for elections to be complete
:param retryWait:
:param customTimeout: specific timeout
:param numInstances: expected number of protocol instances
:return: primary replica for each protocol instance
"""
if retryWait is None:
retryWait = 1
if customTimeout is None:
customTimeout = waits.expectedPoolElectionTimeout(len(nodes))
return checkProtocolInstanceSetup(looper=looper,
nodes=nodes,
retryWait=retryWait,
customTimeout=customTimeout,
numInstances=numInstances)
def test_restart_majority_to_same_view(looper, txnPoolNodeSet, tconf, tdir, allPluginsPath,
sdk_pool_handle, sdk_wallet_client):
# Add transaction to ledger
sdk_send_random_and_check(looper, txnPoolNodeSet, sdk_pool_handle, sdk_wallet_client, 1)
majority = txnPoolNodeSet[:3]
minority = txnPoolNodeSet[3:]
# Restart majority group
tm = tconf.ToleratePrimaryDisconnection + waits.expectedPoolElectionTimeout(len(txnPoolNodeSet))
majority_before_restart = majority.copy()
restart_nodes(looper, txnPoolNodeSet, majority, tconf, tdir, allPluginsPath,
after_restart_timeout=tm, start_one_by_one=False, wait_for_elections=False)
ensureElectionsDone(looper, majority, instances_list=range(2))
# Check that nodes in minority group are aware that they might have inconsistent 3PC state
for node in minority:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 1
# Check that nodes in majority group didn't think they might have inconsistent 3PC state
for node in majority_before_restart:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 0
# Check that nodes in majority group don't think they might have inconsistent 3PC state
for node in majority:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 0
# Restart minority group
restart_nodes(looper, txnPoolNodeSet, minority, tconf, tdir, allPluginsPath,
after_restart_timeout=tm, start_one_by_one=False)
# Check that all nodes are still functional
sdk_ensure_pool_functional(looper, txnPoolNodeSet, sdk_wallet_client, sdk_pool_handle)
def checkProtocolInstanceSetup(looper: Looper,
nodes: Sequence[TestNode],
retryWait: float = 1,
customTimeout: float = None,
numInstances: int = None):
timeout = customTimeout or waits.expectedPoolElectionTimeout(len(nodes))
checkEveryProtocolInstanceHasOnlyOnePrimary(looper=looper,
nodes=nodes,
retryWait=retryWait,
timeout=timeout,
numInstances=numInstances)
checkEveryNodeHasAtMostOnePrimary(looper=looper,
nodes=nodes,
retryWait=retryWait,
customTimeout=timeout)
primaryReplicas = {
replica.instId: replica
for node in nodes for replica in node.replicas if replica.isPrimary
}
return [
r[1]
for r in sorted(primaryReplicas.items(), key=operator.itemgetter(0))
]
def testNodeDoesNotParticipateUntilCaughtUp(txnPoolNodeSet,
nodes_slow_to_process_catchup_reqs,
sdk_node_created_after_some_txns):
"""
A new node that joins after some transactions should stash new transactions
until it has caught up
:return:
"""
looper, new_node, sdk_pool_handle, new_steward_wallet_handle = \
sdk_node_created_after_some_txns
txnPoolNodeSet.append(new_node)
old_nodes = txnPoolNodeSet[:-1]
sdk_send_random_and_check(looper, txnPoolNodeSet, sdk_pool_handle,
new_steward_wallet_handle, 4)
chk_commits_prepares_recvd(0, old_nodes, new_node)
for node in old_nodes:
node.reset_delays_and_process_delayeds()
timeout = waits.expectedPoolCatchupTime(len(txnPoolNodeSet)) + \
catchup_delay + \
waits.expectedPoolElectionTimeout(len(txnPoolNodeSet))
ensureElectionsDone(looper, txnPoolNodeSet, customTimeout=timeout)
waitNodeDataEquality(looper, new_node, *old_nodes,
exclude_from_check=['check_last_ordered_3pc_backup'])
sdk_send_random_and_check(looper, txnPoolNodeSet, sdk_pool_handle,
new_steward_wallet_handle, 2)
# Commits and Prepares are received by all old nodes
with pytest.raises(AssertionError):
# Since nodes discard 3PC messages for already ordered requests.
chk_commits_prepares_recvd(0, old_nodes, new_node)
waitNodeDataEquality(looper, new_node, *old_nodes)
def checkProtocolInstanceSetup(looper: Looper,
nodes: Sequence[TestNode],
retryWait: float = 1,
customTimeout: float = None,
instances: Sequence[int] = None,
check_primaries=True):
timeout = customTimeout or waits.expectedPoolElectionTimeout(len(nodes))
checkEveryProtocolInstanceHasOnlyOnePrimary(looper=looper,
nodes=nodes,
retryWait=retryWait,
timeout=timeout,
instances_list=instances)
checkEveryNodeHasAtMostOnePrimary(looper=looper,
nodes=nodes,
retryWait=retryWait,
customTimeout=timeout)
def check_not_in_view_change():
assert all([not n.master_replica._consensus_data.waiting_for_new_view for n in nodes])
looper.run(eventually(check_not_in_view_change, retryWait=retryWait, timeout=customTimeout))
if check_primaries:
for n in nodes[1:]:
assert nodes[0].primaries == n.primaries
primaryReplicas = {replica.instId: replica
for node in nodes
for replica in node.replicas.values() if replica.isPrimary}
return [r[1] for r in
sorted(primaryReplicas.items(), key=operator.itemgetter(0))]
def checkProtocolInstanceSetup(looper: Looper,
nodes: Sequence[TestNode],
retryWait: float = 1,
customTimeout: float = None,
instances: Sequence[int] = None,
check_primaries=True):
timeout = customTimeout or waits.expectedPoolElectionTimeout(len(nodes))
checkEveryProtocolInstanceHasOnlyOnePrimary(looper=looper,
nodes=nodes,
retryWait=retryWait,
timeout=timeout,
instances_list=instances)
checkEveryNodeHasAtMostOnePrimary(looper=looper,
nodes=nodes,
retryWait=retryWait,
customTimeout=timeout)
if check_primaries:
for n in nodes[1:]:
assert nodes[0].primaries == n.primaries
primaryReplicas = {
replica.instId: replica
for node in nodes for replica in node.replicas.values()
if replica.isPrimary
}
return [
r[1]
for r in sorted(primaryReplicas.items(), key=operator.itemgetter(0))
]
def test_restart_majority_to_same_view(looper, txnPoolNodeSet, tconf, tdir, allPluginsPath,
sdk_pool_handle, sdk_wallet_client):
# Add transaction to ledger
sdk_send_random_and_check(looper, txnPoolNodeSet, sdk_pool_handle, sdk_wallet_client, 1)
majority = txnPoolNodeSet[:3]
minority = txnPoolNodeSet[3:]
# Restart majority group
tm = tconf.ToleratePrimaryDisconnection + waits.expectedPoolElectionTimeout(len(txnPoolNodeSet))
majority_before_restart = majority.copy()
restart_nodes(looper, txnPoolNodeSet, majority, tconf, tdir, allPluginsPath,
after_restart_timeout=tm, start_one_by_one=False, wait_for_elections=False)
ensureElectionsDone(looper, majority, instances_list=range(2))
# Check that nodes in minority group are aware that they might have inconsistent 3PC state
for node in minority:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 1
# Check that nodes in majority group didn't think they might have inconsistent 3PC state
for node in majority_before_restart:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 0
# Check that nodes in majority group don't think they might have inconsistent 3PC state
for node in majority:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 0
# Restart minority group
restart_nodes(looper, txnPoolNodeSet, minority, tconf, tdir, allPluginsPath,
after_restart_timeout=tm, start_one_by_one=False)
# Check that all nodes are still functional
sdk_ensure_pool_functional(looper, txnPoolNodeSet, sdk_wallet_client, sdk_pool_handle)
def testNodeDoesNotParticipateUntilCaughtUp(txnPoolNodeSet,
nodes_slow_to_process_catchup_reqs,
sdk_node_created_after_some_txns):
"""
A new node that joins after some transactions should stash new transactions
until it has caught up
:return:
"""
looper, new_node, sdk_pool_handle, new_steward_wallet_handle = \
sdk_node_created_after_some_txns
txnPoolNodeSet.append(new_node)
old_nodes = txnPoolNodeSet[:-1]
sdk_send_random_and_check(looper, txnPoolNodeSet, sdk_pool_handle,
new_steward_wallet_handle, 5)
chk_commits_prepares_recvd(0, old_nodes, new_node)
for node in old_nodes:
node.reset_delays_and_process_delayeds()
timeout = waits.expectedPoolCatchupTime(len(txnPoolNodeSet)) + \
catchup_delay + \
waits.expectedPoolElectionTimeout(len(txnPoolNodeSet))
ensureElectionsDone(looper, txnPoolNodeSet, customTimeout=timeout)
waitNodeDataEquality(looper, new_node, *old_nodes)
sdk_send_random_and_check(looper, txnPoolNodeSet, sdk_pool_handle,
new_steward_wallet_handle, 2)
# Commits and Prepares are received by all old nodes
with pytest.raises(AssertionError):
# Since nodes discard 3PC messages for already ordered requests.
chk_commits_prepares_recvd(0, old_nodes, new_node)
waitNodeDataEquality(looper, new_node, *old_nodes)
def test_restart_half_to_lower_view(looper, txnPoolNodeSet, tconf, tdir,
allPluginsPath, sdk_pool_handle,
sdk_wallet_client):
# Add transaction to ledger
sdk_send_random_and_check(looper, txnPoolNodeSet, sdk_pool_handle,
sdk_wallet_client, 1)
# Move to higher view
ensure_view_change_complete(looper, txnPoolNodeSet)
# Restart half of nodes
tm = tconf.ToleratePrimaryDisconnection + waits.expectedPoolElectionTimeout(
len(txnPoolNodeSet))
nodes_before_restart = txnPoolNodeSet.copy()
restart_nodes(looper,
txnPoolNodeSet,
txnPoolNodeSet[2:],
tconf,
tdir,
allPluginsPath,
after_restart_timeout=tm,
start_one_by_one=False)
# Check that nodes didn't think they may have inconsistent 3PC state
for node in nodes_before_restart:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 0
# Check that nodes don't think they may have inconsistent 3PC state
for node in txnPoolNodeSet:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 0
# Check that all nodes are still functional
sdk_ensure_pool_functional(looper, txnPoolNodeSet, sdk_wallet_client,
sdk_pool_handle)
def ensureElectionsDone(looper: Looper,
nodes: Sequence[TestNode],
retryWait: float = None, # seconds
customTimeout: float = None,
instances_list: Sequence[int] = None) -> Sequence[TestNode]:
# TODO: Change the name to something like `ensure_primaries_selected`
# since there might not always be an election, there might be a round
# robin selection
"""
Wait for elections to be complete
:param retryWait:
:param customTimeout: specific timeout
:param instances_list: expected number of protocol instances
:return: primary replica for each protocol instance
"""
if retryWait is None:
retryWait = 1
if customTimeout is None:
customTimeout = waits.expectedPoolElectionTimeout(len(nodes))
return checkProtocolInstanceSetup(
looper=looper,
nodes=nodes,
retryWait=retryWait,
customTimeout=customTimeout,
instances=instances_list)
def testNodeDoesNotParticipateUntilCaughtUp(txnPoolNodeSet,
nodes_slow_to_process_catchup_reqs,
nodeCreatedAfterSomeTxns):
"""
A new node that joins after some transactions should stash new transactions
until it has caught up
:return:
"""
looper, new_node, client, wallet, newStewardClient, newStewardWallet = \
nodeCreatedAfterSomeTxns
txnPoolNodeSet.append(new_node)
old_nodes = txnPoolNodeSet[:-1]
sendReqsToNodesAndVerifySuffReplies(looper, wallet, client, 5)
chk_commits_prepares_recvd(0, old_nodes, new_node)
for node in old_nodes:
node.reset_delays_and_process_delayeds()
timeout = waits.expectedPoolCatchupTime(len(txnPoolNodeSet)) + \
catchup_delay + \
waits.expectedPoolElectionTimeout(len(txnPoolNodeSet))
ensureElectionsDone(looper, txnPoolNodeSet, customTimeout=timeout)
waitNodeDataEquality(looper, new_node, *old_nodes)
sendReqsToNodesAndVerifySuffReplies(looper, wallet, client, 2)
# Commits and Prepares are received by all old nodes
with pytest.raises(AssertionError):
# Since nodes discard 3PC messages for already ordered requests.
chk_commits_prepares_recvd(0, old_nodes, new_node)
waitNodeDataEquality(looper, new_node, *old_nodes)
def changeNodeHa(looper, txnPoolNodeSet, tdirWithClientPoolTxns,
poolTxnData, poolTxnStewardNames, tconf, shouldBePrimary, tdir):
# prepare new ha for node and client stack
subjectedNode = None
stewardName = None
stewardsSeed = None
for nodeIndex, n in enumerate(txnPoolNodeSet):
if shouldBePrimary == n.has_master_primary:
subjectedNode = n
stewardName = poolTxnStewardNames[nodeIndex]
stewardsSeed = poolTxnData["seeds"][stewardName].encode()
break
nodeStackNewHA, clientStackNewHA = genHa(2)
logger.debug("change HA for node: {} to {}".format(
subjectedNode.name, (nodeStackNewHA, clientStackNewHA)))
nodeSeed = poolTxnData["seeds"][subjectedNode.name].encode()
# change HA
stewardClient, req = changeHA(looper, tconf, subjectedNode.name, nodeSeed,
nodeStackNewHA, stewardName, stewardsSeed,
basedir=tdirWithClientPoolTxns)
waitForSufficientRepliesForRequests(looper, stewardClient,
requests=[req])
# stop node for which HA will be changed
subjectedNode.stop()
looper.removeProdable(subjectedNode)
# start node with new HA
config_helper = PNodeConfigHelper(subjectedNode.name, tconf, chroot=tdir)
restartedNode = TestNode(subjectedNode.name,
config_helper=config_helper,
config=tconf, ha=nodeStackNewHA,
cliha=clientStackNewHA)
looper.add(restartedNode)
txnPoolNodeSet[nodeIndex] = restartedNode
looper.run(checkNodesConnected(txnPoolNodeSet, customTimeout=70))
electionTimeout = waits.expectedPoolElectionTimeout(
nodeCount=len(txnPoolNodeSet),
numOfReelections=3)
ensureElectionsDone(looper,
txnPoolNodeSet,
retryWait=1,
customTimeout=electionTimeout)
# start client and check the node HA
anotherClient, _ = genTestClient(tmpdir=tdirWithClientPoolTxns,
usePoolLedger=True)
looper.add(anotherClient)
looper.run(eventually(anotherClient.ensureConnectedToNodes))
stewardWallet = Wallet(stewardName)
stewardWallet.addIdentifier(signer=DidSigner(seed=stewardsSeed))
sendReqsToNodesAndVerifySuffReplies(
looper, stewardWallet, stewardClient, 8)
def test_all_replicas_hold_request_keys(
perf_chk_patched,
looper,
txnPoolNodeSet,
sdk_wallet_client,
sdk_pool_handle):
"""
All replicas whether primary or non primary hold request keys of forwarded
requests. Once requests are ordered, they request keys are removed from replica.
"""
tconf = perf_chk_patched
delay_3pc = 2
delay_3pc_messages(txnPoolNodeSet, 0, delay_3pc)
delay_3pc_messages(txnPoolNodeSet, 1, delay_3pc)
def chk(count):
# All replicas have same amount of forwarded request keys and all keys
# are finalised.
for node in txnPoolNodeSet:
for r in node.replicas.values():
if r.isPrimary is False:
assert len(r.requestQueues[DOMAIN_LEDGER_ID]) == count
for i in range(count):
k = r.requestQueues[DOMAIN_LEDGER_ID][i]
assert r.requests[k].finalised
elif r.isPrimary is True:
assert len(r.requestQueues[DOMAIN_LEDGER_ID]) == 0
reqs = sdk_signed_random_requests(looper,
sdk_wallet_client,
tconf.Max3PCBatchSize - 1)
req_resps = sdk_send_signed_requests(sdk_pool_handle, reqs)
# Only non primary replicas should have all request keys with them
looper.run(eventually(chk, tconf.Max3PCBatchSize - 1))
sdk_get_replies(looper, req_resps, timeout=sdk_eval_timeout(
tconf.Max3PCBatchSize - 1, len(txnPoolNodeSet),
add_delay_to_timeout=delay_3pc))
# Replicas should have no request keys with them since they are ordered
looper.run(eventually(chk, 0)) # Need to wait since one node might not
# have processed it.
delay = 1
for node in txnPoolNodeSet:
node.nodeIbStasher.delay(nom_delay(delay))
ensure_view_change(looper, txnPoolNodeSet)
reqs = sdk_signed_random_requests(looper,
sdk_wallet_client,
2 * tconf.Max3PCBatchSize)
req_resps = sdk_send_signed_requests(sdk_pool_handle, reqs)
looper.run(eventually(chk, 2 * tconf.Max3PCBatchSize))
# Since each nomination is delayed and there will be multiple nominations
# so adding some extra time
timeout = waits.expectedPoolElectionTimeout(len(txnPoolNodeSet)) + \
len(txnPoolNodeSet) * delay
ensureElectionsDone(looper, txnPoolNodeSet, customTimeout=timeout)
sdk_get_replies(looper, req_resps, timeout=timeout)
looper.run(eventually(chk, 0))
def test_pool_reaches_quorum_after_f_plus_2_nodes_turned_off_and_later_on(
looper, allPluginsPath, tdir, tconf, txnPoolNodeSet, wallet1, client1,
client1Connected):
nodes = txnPoolNodeSet
initial_view_no = nodes[0].viewNo
request = sendRandomRequest(wallet1, client1)
waitForSufficientRepliesForRequests(looper, client1, requests=[request])
stop_node(nodes[0], looper, nodes)
waitForViewChange(looper, nodes[1:], expectedViewNo=initial_view_no + 1)
ensureElectionsDone(looper,
nodes[1:],
numInstances=getRequiredInstances(nodeCount))
request = sendRandomRequest(wallet1, client1)
waitForSufficientRepliesForRequests(looper, client1, requests=[request])
stop_node(nodes[1], looper, nodes)
looper.runFor(tconf.ToleratePrimaryDisconnection + 2)
checkViewNoForNodes(nodes[2:], initial_view_no + 1)
request = sendRandomRequest(wallet1, client1)
verify_request_not_replied_and_not_ordered(request, looper, client1, nodes)
stop_node(nodes[2], looper, nodes)
looper.runFor(tconf.ToleratePrimaryDisconnection + 2)
checkViewNoForNodes(nodes[3:], initial_view_no + 1)
request = sendRandomRequest(wallet1, client1)
verify_request_not_replied_and_not_ordered(request, looper, client1, nodes)
nodes[2] = start_stopped_node(nodes[2], looper, tconf, tdir,
allPluginsPath)
looper.runFor(waits.expectedPoolElectionTimeout(len(nodes)))
request = sendRandomRequest(wallet1, client1)
verify_request_not_replied_and_not_ordered(request, looper, client1, nodes)
nodes[1] = start_stopped_node(nodes[1], looper, tconf, tdir,
allPluginsPath)
ensureElectionsDone(looper,
nodes[1:],
numInstances=getRequiredInstances(nodeCount))
waitForViewChange(looper, nodes[1:], expectedViewNo=initial_view_no + 1)
request = sendRandomRequest(wallet1, client1)
waitForSufficientRepliesForRequests(looper, client1, requests=[request])
nodes[0] = start_stopped_node(nodes[0], looper, tconf, tdir,
allPluginsPath)
ensureElectionsDone(looper,
nodes,
numInstances=getRequiredInstances(nodeCount))
waitForViewChange(looper, nodes, expectedViewNo=initial_view_no + 1)
request = sendRandomRequest(wallet1, client1)
waitForSufficientRepliesForRequests(looper, client1, requests=[request])
def changeNodeHa(looper, txnPoolNodeSet, tdirWithClientPoolTxns,
tconf, shouldBePrimary, tdir,
sdk_pool_handle, sdk_wallet_stewards):
# prepare new ha for node and client stack
subjectedNode = None
node_index = None
for nodeIndex, n in enumerate(txnPoolNodeSet):
if shouldBePrimary == n.has_master_primary:
subjectedNode = n
node_index = nodeIndex
break
nodeStackNewHA, clientStackNewHA = genHa(2)
logger.debug("change HA for node: {} to {}".format(
subjectedNode.name, (nodeStackNewHA, clientStackNewHA)))
# change HA
sdk_wallet_steward = sdk_wallet_stewards[node_index]
node_dest = hexToFriendly(subjectedNode.nodestack.verhex)
sdk_send_update_node(looper, sdk_wallet_steward,
sdk_pool_handle,
node_dest, subjectedNode.name,
nodeStackNewHA[0], nodeStackNewHA[1],
clientStackNewHA[0], clientStackNewHA[1],
services=[VALIDATOR])
# stop node for which HA will be changed
subjectedNode.stop()
looper.removeProdable(subjectedNode)
# start node with new HA
config_helper = PNodeConfigHelper(subjectedNode.name, tconf, chroot=tdir)
restartedNode = TestNode(subjectedNode.name,
config_helper=config_helper,
config=tconf, ha=nodeStackNewHA,
cliha=clientStackNewHA)
looper.add(restartedNode)
txnPoolNodeSet[nodeIndex] = restartedNode
looper.run(checkNodesConnected(txnPoolNodeSet, customTimeout=70))
sdk_pool_refresh(looper, sdk_pool_handle)
electionTimeout = waits.expectedPoolElectionTimeout(
nodeCount=len(txnPoolNodeSet),
numOfReelections=3)
ensureElectionsDone(looper,
txnPoolNodeSet,
retryWait=1,
customTimeout=electionTimeout)
# start client and check the node HA
anotherClient, _ = genTestClient(tmpdir=tdirWithClientPoolTxns,
usePoolLedger=True)
looper.add(anotherClient)
looper.run(eventually(anotherClient.ensureConnectedToNodes))
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle,
sdk_wallet_steward,
8)
def test_restart_groups_wp(looper, txnPoolNodeSet, tconf, tdir,
sdk_pool_handle, sdk_wallet_client, allPluginsPath):
tm = tconf.ToleratePrimaryDisconnection + waits.expectedPoolElectionTimeout(len(txnPoolNodeSet))
restart_group = get_group(txnPoolNodeSet, 7, include_primary=True)
restart_nodes(looper, txnPoolNodeSet, restart_group, tconf, tdir, allPluginsPath,
after_restart_timeout=tm, per_add_timeout=None)
sdk_ensure_pool_functional(looper, txnPoolNodeSet, sdk_wallet_client, sdk_pool_handle)
def test_restart_groups_6_of_7_np_no_tm(looper, txnPoolNodeSet, tconf, tdir,
sdk_pool_handle, sdk_wallet_client, allPluginsPath):
tm = tconf.ToleratePrimaryDisconnection + waits.expectedPoolElectionTimeout(len(txnPoolNodeSet))
restart_group = get_group(txnPoolNodeSet, 6, include_primary=False)
restart_nodes(looper, txnPoolNodeSet, restart_group, tconf, tdir, allPluginsPath,
after_restart_timeout=tm, start_one_by_one=False)
sdk_ensure_pool_functional(looper, txnPoolNodeSet, sdk_wallet_client, sdk_pool_handle)
def changeNodeHa(looper, txnPoolNodeSet,
tconf, shouldBePrimary, tdir,
sdk_pool_handle, sdk_wallet_stewards,
sdk_wallet_client):
# prepare new ha for node and client stack
subjectedNode = None
node_index = None
for nodeIndex, n in enumerate(txnPoolNodeSet):
if shouldBePrimary == n.has_master_primary:
subjectedNode = n
node_index = nodeIndex
break
nodeStackNewHA, clientStackNewHA = genHa(2)
logger.debug("change HA for node: {} to {}".format(
subjectedNode.name, (nodeStackNewHA, clientStackNewHA)))
# change HA
sdk_wallet_steward = sdk_wallet_stewards[node_index]
node_dest = hexToFriendly(subjectedNode.nodestack.verhex)
sdk_send_update_node(looper, sdk_wallet_steward,
sdk_pool_handle,
node_dest, subjectedNode.name,
nodeStackNewHA[0], nodeStackNewHA[1],
clientStackNewHA[0], clientStackNewHA[1],
services=[VALIDATOR])
# stop node for which HA will be changed
subjectedNode.stop()
looper.removeProdable(subjectedNode)
# start node with new HA
config_helper = PNodeConfigHelper(subjectedNode.name, tconf, chroot=tdir)
restartedNode = TestNode(subjectedNode.name,
config_helper=config_helper,
config=tconf, ha=nodeStackNewHA,
cliha=clientStackNewHA)
looper.add(restartedNode)
txnPoolNodeSet[nodeIndex] = restartedNode
looper.run(checkNodesConnected(txnPoolNodeSet, customTimeout=70))
electionTimeout = waits.expectedPoolElectionTimeout(
nodeCount=len(txnPoolNodeSet),
numOfReelections=3)
ensureElectionsDone(looper,
txnPoolNodeSet,
retryWait=1,
customTimeout=electionTimeout)
sdk_pool_refresh(looper, sdk_pool_handle)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle,
sdk_wallet_client,
8)
def test_all_replicas_hold_request_keys(perf_chk_patched, looper,
txnPoolNodeSet, sdk_wallet_client,
sdk_pool_handle):
"""
All replicas whether primary or non primary hold request keys of forwarded
requests. Once requests are ordered, they request keys are removed from replica.
"""
tconf = perf_chk_patched
delay_3pc = 2
delay_3pc_messages(txnPoolNodeSet, 0, delay_3pc)
delay_3pc_messages(txnPoolNodeSet, 1, delay_3pc)
def chk(count):
# All replicas have same amount of forwarded request keys and all keys
# are finalised.
for node in txnPoolNodeSet:
for r in node.replicas.values():
if r.isPrimary is False:
assert len(r.requestQueues[DOMAIN_LEDGER_ID]) == count
for i in range(count):
k = r.requestQueues[DOMAIN_LEDGER_ID][i]
assert r.requests[k].finalised
elif r.isPrimary is True:
assert len(r.requestQueues[DOMAIN_LEDGER_ID]) == 0
reqs = sdk_signed_random_requests(looper, sdk_wallet_client,
tconf.Max3PCBatchSize - 1)
req_resps = sdk_send_signed_requests(sdk_pool_handle, reqs)
# Only non primary replicas should have all request keys with them
looper.run(eventually(chk, tconf.Max3PCBatchSize - 1))
sdk_get_replies(looper,
req_resps,
timeout=sdk_eval_timeout(tconf.Max3PCBatchSize - 1,
len(txnPoolNodeSet),
add_delay_to_timeout=delay_3pc))
# Replicas should have no request keys with them since they are ordered
looper.run(eventually(chk, 0)) # Need to wait since one node might not
# have processed it.
delay = 1
for node in txnPoolNodeSet:
node.nodeIbStasher.delay(nom_delay(delay))
ensure_view_change(looper, txnPoolNodeSet)
reqs = sdk_signed_random_requests(looper, sdk_wallet_client,
2 * tconf.Max3PCBatchSize)
req_resps = sdk_send_signed_requests(sdk_pool_handle, reqs)
looper.run(eventually(chk, 2 * tconf.Max3PCBatchSize))
# Since each nomination is delayed and there will be multiple nominations
# so adding some extra time
timeout = waits.expectedPoolElectionTimeout(len(txnPoolNodeSet)) + \
len(txnPoolNodeSet) * delay
ensureElectionsDone(looper, txnPoolNodeSet, customTimeout=timeout)
sdk_get_replies(looper, req_resps, timeout=timeout)
looper.run(eventually(chk, 0))
def waitForViewChange(looper, nodeSet, expectedViewNo=None, customTimeout=None):
"""
Waits for nodes to come to same view.
Raises exception when time is out
"""
timeout = customTimeout or waits.expectedPoolElectionTimeout(len(nodeSet))
return looper.run(eventually(checkViewNoForNodes,
nodeSet,
expectedViewNo,
timeout=timeout))
def test_view_change_without_primary(nodeSet, looper,
patched_view_change_timeout):
first, others = stop_nodes_and_remove_first(looper, nodeSet)
start_and_connect_nodes(looper, others)
timeout = waits.expectedPoolElectionTimeout(len(nodeSet)) + patched_view_change_timeout
checkProtocolInstanceSetup(looper=looper, nodes=others, retryWait=1,
customTimeout=timeout,
numInstances=getRequiredInstances(len(nodeSet)))
def waitForViewChange(looper, txnPoolNodeSet, expectedViewNo=None,
customTimeout=None):
"""
Waits for nodes to come to same view.
Raises exception when time is out
"""
timeout = customTimeout or waits.expectedPoolElectionTimeout(len(txnPoolNodeSet))
return looper.run(eventually(checkViewNoForNodes,
txnPoolNodeSet,
expectedViewNo,
timeout=timeout))
def testPrimaryElectionCase1(case1Setup, looper, keySharedNodes):
"""
Case 1 - A node making multiple nominations for a particular node. Consider
4 nodes A, B, C and D. Lets say node B is malicious and is repeatedly
nominating Node D
"""
nodes = keySharedNodes
nodeA, nodeB, nodeC, nodeD = [nodes.getNode(nm) for nm in nodes.nodeNames]
# Doesn't matter if nodes reach the ready state or not. Just start them
looper.run(checkNodesConnected(nodes))
# Node B sends multiple NOMINATE messages for Node D but only after A has
# nominated itself
timeout = waits.expectedPoolNominationTimeout(
nodeCount=len(keySharedNodes))
looper.run(
eventually(checkNomination,
nodeA,
nodeA.name,
retryWait=.25,
timeout=timeout))
instId = getSelfNominationByNode(nodeA)
for i in range(5):
# nodeB.send(Nomination(nodeD.name, instId, nodeB.viewNo))
nodeB.send(nominationByNode(nodeD.name, nodeB, instId))
nodeB.nodestack.flushOutBoxes()
# No node from node A, node C, node D(node B is malicious anyway so not
# considering it) should have more than one nomination for node D since
# node D is slow. The one nomination for D, that nodes A, C
# and D might have would be because of node B
for node in [nodeA, nodeC, nodeD]:
assert [n[0] for n in node.elector.nominations[instId].values()].count(
Replica.generateName(nodeD.name, instId)) \
<= 1
timeout = waits.expectedPoolElectionTimeout(nodeCount) + delayOfNomination
primaryReplicas = ensureElectionsDone(looper=looper,
nodes=nodes,
customTimeout=timeout)
for node in nodes:
logger.debug("{}'s nominations {}".format(node,
node.elector.nominations))
# Node D should not have any primary
assert not nodeD.hasPrimary
# A node other than Node D should not have any replica among the
# primary replicas
assert nodeD.name not in [pr.name for pr in primaryReplicas]
def test_view_change_without_primary(txnPoolNodeSet, looper,
patched_view_change_timeout):
first, others = stop_nodes_and_remove_first(looper, txnPoolNodeSet)
start_and_connect_nodes(looper, others)
timeout = waits.expectedPoolElectionTimeout(len(txnPoolNodeSet)) + patched_view_change_timeout
#looper.runFor(40)
checkProtocolInstanceSetup(looper=looper, nodes=txnPoolNodeSet, retryWait=1,
customTimeout=timeout,
instances=range(getRequiredInstances(len(txnPoolNodeSet))))
def test_all_replicas_hold_request_keys(looper, txnPoolNodeSet, client1,
wallet1, client1Connected, tconf):
"""
All replicas whether primary or non primary hold request keys of forwarded
requests. Once requests are ordered, they request keys are removed from replica.
"""
delay_3pc_messages(txnPoolNodeSet, 0, 2)
delay_3pc_messages(txnPoolNodeSet, 1, 2)
def chk(count):
# All replicas have same amount of forwarded request keys and all keys
# are finalised.
for node in txnPoolNodeSet:
for r in node.replicas:
if r.isPrimary is False:
assert len(r.requestQueues[DOMAIN_LEDGER_ID]) == count
for i in range(count):
k = r.requestQueues[DOMAIN_LEDGER_ID][i]
assert r.requests[k].finalised
elif r.isPrimary is True:
assert len(r.requestQueues[DOMAIN_LEDGER_ID]) == 0
reqs = sendRandomRequests(wallet1, client1, tconf.Max3PCBatchSize - 1)
# Only non primary replicas should have all request keys with them
looper.run(eventually(chk, tconf.Max3PCBatchSize - 1))
waitForSufficientRepliesForRequests(looper,
client1,
requests=reqs,
add_delay_to_timeout=2)
# Replicas should have no request keys with them since they are ordered
looper.run(eventually(chk, 0)) # Need to wait since one node might not
# have processed it.
delay = 1
for node in txnPoolNodeSet:
node.nodeIbStasher.delay(nom_delay(delay))
ensure_view_change(looper, txnPoolNodeSet)
reqs = sendRandomRequests(wallet1, client1, 2 * tconf.Max3PCBatchSize)
looper.run(eventually(chk, 2 * tconf.Max3PCBatchSize))
# Since each nomination is delayed and there will be multiple nominations
# so adding some extra time
timeout = waits.expectedPoolElectionTimeout(len(txnPoolNodeSet)) + \
len(txnPoolNodeSet)*delay
ensureElectionsDone(looper, txnPoolNodeSet, customTimeout=timeout)
waitForSufficientRepliesForRequests(looper,
client1,
requests=reqs,
add_delay_to_timeout=2)
looper.run(eventually(chk, 0))
def checkEveryNodeHasAtMostOnePrimary(looper: Looper,
nodes: Sequence[TestNode],
retryWait: float = None,
customTimeout: float = None):
def checkAtMostOnePrim(node):
prims = [r for r in node.replicas if r.isPrimary]
assert len(prims) <= 1
timeout = customTimeout or waits.expectedPoolElectionTimeout(len(nodes))
for node in nodes:
looper.run(eventually(checkAtMostOnePrim,
node,
retryWait=retryWait,
timeout=timeout))
def checkEveryNodeHasAtMostOnePrimary(looper: Looper,
nodes: Sequence[TestNode],
retryWait: float = None,
customTimeout: float = None):
def checkAtMostOnePrim(node):
prims = [r for r in node.replicas.values() if r.isPrimary]
assert len(prims) <= 1
timeout = customTimeout or waits.expectedPoolElectionTimeout(len(nodes))
for node in nodes:
looper.run(eventually(checkAtMostOnePrim,
node,
retryWait=retryWait,
timeout=timeout))
def test_view_change_without_primary(txnPoolNodeSet, looper, tconf):
first, others = stop_nodes_and_remove_first(looper, txnPoolNodeSet)
start_and_connect_nodes(looper, others)
timeout = waits.expectedPoolElectionTimeout(
len(txnPoolNodeSet)) + tconf.NEW_VIEW_TIMEOUT
# looper.runFor(40)
checkProtocolInstanceSetup(looper=looper,
nodes=txnPoolNodeSet,
retryWait=1,
customTimeout=timeout,
instances=range(
getRequiredInstances(len(txnPoolNodeSet))))
def test_restart_to_same_view_with_killed_primary(looper, txnPoolNodeSet, tconf, tdir, allPluginsPath,
sdk_pool_handle, sdk_wallet_client):
restart_timeout = tconf.ToleratePrimaryDisconnection + \
waits.expectedPoolElectionTimeout(len(txnPoolNodeSet))
primary = txnPoolNodeSet[0]
alive_nodes = txnPoolNodeSet[1:]
minority = alive_nodes[-1:]
majority = alive_nodes[:-1]
# Move to higher view by killing primary
primary.cleanupOnStopping = True
primary.stop()
looper.removeProdable(primary)
ensure_node_disconnected(looper, primary, txnPoolNodeSet)
waitForViewChange(looper, alive_nodes, 1, customTimeout=VIEW_CHANGE_TIMEOUT)
ensureElectionsDone(looper, alive_nodes, instances_list=range(3))
# Add transaction to ledger
sdk_send_random_and_check(looper, alive_nodes, sdk_pool_handle, sdk_wallet_client, 1)
# Restart majority group
majority_before_restart = majority.copy()
restart_nodes(looper, alive_nodes, majority, tconf, tdir, allPluginsPath,
after_restart_timeout=restart_timeout, start_one_by_one=False, wait_for_elections=False)
waitForViewChange(looper, majority, 1, customTimeout=2.1 * VIEW_CHANGE_TIMEOUT)
ensureElectionsDone(looper, majority, instances_list=range(3))
# Check that nodes in minority group are aware that they might have inconsistent 3PC state
for node in minority:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 1
# Check that nodes in majority group didn't think they might have inconsistent 3PC state
for node in majority_before_restart:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 0
# Check that nodes in majority group don't think they might have inconsistent 3PC state
for node in majority:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 0
# Restart minority group
restart_nodes(looper, alive_nodes, minority, tconf, tdir, allPluginsPath,
after_restart_timeout=restart_timeout, start_one_by_one=False, wait_for_elections=False)
ensureElectionsDone(looper, alive_nodes, instances_list=range(3))
# Check that all nodes are still functional
sdk_ensure_pool_functional(looper, alive_nodes, sdk_wallet_client, sdk_pool_handle)
def test_restart_to_same_view_with_killed_primary(looper, txnPoolNodeSet, tconf, tdir, allPluginsPath,
sdk_pool_handle, sdk_wallet_client):
restart_timeout = tconf.ToleratePrimaryDisconnection + \
waits.expectedPoolElectionTimeout(len(txnPoolNodeSet))
primary = txnPoolNodeSet[0]
alive_nodes = txnPoolNodeSet[1:]
minority = alive_nodes[-1:]
majority = alive_nodes[:-1]
# Move to higher view by killing primary
primary.cleanupOnStopping = True
primary.stop()
looper.removeProdable(primary)
ensure_node_disconnected(looper, primary, txnPoolNodeSet)
waitForViewChange(looper, alive_nodes, 1, customTimeout=VIEW_CHANGE_TIMEOUT)
ensureElectionsDone(looper, alive_nodes, numInstances=3)
# Add transaction to ledger
sdk_send_random_and_check(looper, alive_nodes, sdk_pool_handle, sdk_wallet_client, 1)
# Restart majority group
majority_before_restart = majority.copy()
restart_nodes(looper, alive_nodes, majority, tconf, tdir, allPluginsPath,
after_restart_timeout=restart_timeout, start_one_by_one=False, wait_for_elections=False)
waitForViewChange(looper, majority, 1, customTimeout=2.1 * VIEW_CHANGE_TIMEOUT)
ensureElectionsDone(looper, majority, numInstances=3)
# Check that nodes in minority group are aware that they might have inconsistent 3PC state
for node in minority:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 1
# Check that nodes in majority group didn't think they might have inconsistent 3PC state
for node in majority_before_restart:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 0
# Check that nodes in majority group don't think they might have inconsistent 3PC state
for node in majority:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 0
# Restart minority group
restart_nodes(looper, alive_nodes, minority, tconf, tdir, allPluginsPath,
after_restart_timeout=restart_timeout, start_one_by_one=False, wait_for_elections=False)
ensureElectionsDone(looper, alive_nodes, numInstances=3)
# Check that all nodes are still functional
sdk_ensure_pool_functional(looper, alive_nodes, sdk_wallet_client, sdk_pool_handle)
def testPrimaryElectionCase1(case1Setup, looper, txnPoolNodeSet):
"""
Case 1 - A node making multiple nominations for a particular node. Consider
4 nodes A, B, C and D. Lets say node B is malicious and is repeatedly
nominating Node D
"""
nodeA, nodeB, nodeC, nodeD = txnPoolNodeSet
# Doesn't matter if nodes reach the ready state or not. Just start them
looper.run(checkNodesConnected(txnPoolNodeSet))
# Node B sends multiple NOMINATE messages for Node D but only after A has
# nominated itself
timeout = waits.expectedPoolNominationTimeout(
nodeCount=len(txnPoolNodeSet))
looper.run(eventually(checkNomination, nodeA, nodeA.name,
retryWait=.25, timeout=timeout))
instId = getSelfNominationByNode(nodeA)
for i in range(5):
# nodeB.send(Nomination(nodeD.name, instId, nodeB.viewNo))
nodeB.send(nominationByNode(nodeD.name, nodeB, instId))
nodeB.nodestack.flushOutBoxes()
# No node from node A, node C, node D(node B is malicious anyway so not
# considering it) should have more than one nomination for node D since
# node D is slow. The one nomination for D, that nodes A, C
# and D might have would be because of node B
for node in [nodeA, nodeC, nodeD]:
assert [n[0] for n in node.elector.nominations[instId].values()].count(
Replica.generateName(nodeD.name, instId)) \
<= 1
timeout = waits.expectedPoolElectionTimeout(nodeCount) + delayOfNomination
primaryReplicas = ensureElectionsDone(looper=looper,
nodes=txnPoolNodeSet, customTimeout=timeout)
for node in txnPoolNodeSet:
logger.debug(
"{}'s nominations {}".format(node, node.elector.nominations))
# Node D should not have any primary
assert not nodeD.hasPrimary
# A node other than Node D should not have any replica among the
# primary replicas
assert nodeD.name not in [pr.name for pr in primaryReplicas]
def testPrimarySelectionAfterPoolReady(
looper, txnPoolNodeSet, sdk_pool_handle, sdk_wallet_steward):
"""
Once the pool is ready(node has connected to at least 3 other nodes),
appropriate primary replicas should be selected.
"""
def checkPrimaryPlacement():
# Node names sorted by rank
sortedNodes = sorted(txnPoolNodeSet,
key=operator.attrgetter("rank"))
for idx, node in enumerate(sortedNodes):
# For instance 0, the primary replica should be on the node with
# rank 0
if idx == 0:
Replica.generateName(sortedNodes[idx].name, 0)
assert node.replicas[0].isPrimary
assert not node.replicas[1].isPrimary
assert not node.replicas[2].isPrimary
# For instance 1, the primary replica should be on the node with
# rank 1
if idx == 1:
Replica.generateName(sortedNodes[idx].name, 1)
assert not node.replicas[0].isPrimary
assert node.replicas[1].isPrimary
assert not node.replicas[2].isPrimary
# For instance 2, the primary replica should be on the node with
# rank 2
if idx == 2:
Replica.generateName(sortedNodes[idx].name, 2)
assert not node.replicas[0].isPrimary
assert not node.replicas[1].isPrimary
assert node.replicas[2].isPrimary
check_rank_consistent_across_each_node(txnPoolNodeSet)
# Check if the primary is on the correct node
timeout = waits.expectedPoolElectionTimeout(len(txnPoolNodeSet))
looper.run(eventually(checkPrimaryPlacement, retryWait=1, timeout=timeout))
# Check if every protocol instance has one and only one primary and any node
# has no more than one primary
checkProtocolInstanceSetup(looper, txnPoolNodeSet, retryWait=1)
sdk_send_random_and_check(looper, txnPoolNodeSet, sdk_pool_handle,
sdk_wallet_steward, 5)
def testPrimarySelectionAfterPoolReady(
looper, txnPoolNodeSet, sdk_pool_handle, sdk_wallet_steward):
"""
Once the pool is ready(node has connected to at least 3 other nodes),
appropriate primary replicas should be selected.
"""
def checkPrimaryPlacement():
# Node names sorted by rank
sortedNodes = sorted(txnPoolNodeSet,
key=operator.attrgetter("rank"))
for idx, node in enumerate(sortedNodes):
# For instance 0, the primary replica should be on the node with
# rank 0
if idx == 0:
Replica.generateName(sortedNodes[idx].name, 0)
assert node.replicas[0].isPrimary
assert not node.replicas[1].isPrimary
assert not node.replicas[2].isPrimary
# For instance 1, the primary replica should be on the node with
# rank 1
if idx == 1:
Replica.generateName(sortedNodes[idx].name, 1)
assert not node.replicas[0].isPrimary
assert node.replicas[1].isPrimary
assert not node.replicas[2].isPrimary
# For instance 2, the primary replica should be on the node with
# rank 2
if idx == 2:
Replica.generateName(sortedNodes[idx].name, 2)
assert not node.replicas[0].isPrimary
assert not node.replicas[1].isPrimary
assert node.replicas[2].isPrimary
check_rank_consistent_across_each_node(txnPoolNodeSet)
# Check if the primary is on the correct node
timeout = waits.expectedPoolElectionTimeout(len(txnPoolNodeSet))
looper.run(eventually(checkPrimaryPlacement, retryWait=1, timeout=timeout))
# Check if every protocol instance has one and only one primary and any node
# has no more than one primary
checkProtocolInstanceSetup(looper, txnPoolNodeSet, retryWait=1)
sdk_send_random_and_check(looper, txnPoolNodeSet, sdk_pool_handle,
sdk_wallet_steward, 5)
def testPrimaryElectionCase5(case5Setup, looper, txnPoolNodeSet):
"""
Case 5 - A node making primary declarations for a multiple other nodes.
Consider 4 nodes A, B, C, and D. Lets say node B is malicious and
declares node C as primary to all nodes.
Again node B declares node D as primary to all nodes.
"""
A, B, C, D = txnPoolNodeSet
looper.run(checkNodesConnected(txnPoolNodeSet))
BRep = Replica.generateName(B.name, 0)
CRep = Replica.generateName(C.name, 0)
DRep = Replica.generateName(D.name, 0)
# Node B first sends PRIMARY msgs for Node C to all nodes
# B.send(Primary(CRep, 0, B.viewNo))
B.send(primaryByNode(CRep, B, 0))
# Node B sends PRIMARY msgs for Node D to all nodes
# B.send(Primary(DRep, 0, B.viewNo))
B.send(primaryByNode(DRep, B, 0))
# Ensure elections are done
# also have to take into account the catchup procedure
timeout = waits.expectedPoolElectionTimeout(len(txnPoolNodeSet)) + \
waits.expectedPoolCatchupTime(len(txnPoolNodeSet)) + \
delayOfElectionDone
ensureElectionsDone(looper=looper,
nodes=txnPoolNodeSet,
customTimeout=timeout)
# All nodes from node A, node C, node D(node B is malicious anyway so not
# considering it) should have primary declarations for node C from node B
# since node B first nominated node C
for node in [A, C, D]:
logger.debug(
"node {} should have primary declaration for C from node B".format(
node))
assert node.elector.primaryDeclarations[0][BRep][0] == CRep
def testPrimaryElectionCase4(case4Setup, looper):
"""
Case 4 - A node making multiple primary declarations for a particular node.
Consider 4 nodes A, B, C and D. Lets say node B is malicious and is
repeatedly declaring Node D as primary
"""
allNodes = case4Setup
A, B, C, D = allNodes
looper.run(checkNodesConnected(allNodes))
# Node B sends multiple declarations of node D's 0th protocol instance as
# primary to all nodes
for i in range(5):
# B.send(Primary(D.name, 0, B.viewNo))
B.send(primaryByNode(D.name, B, 0))
# No node from node A, node C, node D(node B is malicious anyway so not
# considering it) should have more than one primary declaration for node
# D since node D is slow. The one primary declaration for node D,
# that nodes A, C and D might have would be because of node B
def x():
primDecs = [p[0] for p in node.elector.primaryDeclarations[0].values()]
assert primDecs.count(D.name) <= 1
# also have to take into account the catchup procedure
timeout = waits.expectedPoolNominationTimeout(len(allNodes)) + \
waits.expectedPoolCatchupTime(len(allNodes))
for node in (A, C, D):
looper.run(eventually(x, retryWait=.5, timeout=timeout))
timeout = waits.expectedPoolElectionTimeout(
len(allNodes)) + delaySelfNomination
ensureElectionsDone(looper=looper, nodes=allNodes, customTimeout=timeout)
# Node D should not have any primary replica
assert not D.hasPrimary
def testPrimaryElectionCase4(case4Setup, looper):
"""
Case 4 - A node making multiple primary declarations for a particular node.
Consider 4 nodes A, B, C and D. Lets say node B is malicious and is
repeatedly declaring Node D as primary
"""
allNodes = case4Setup
A, B, C, D = allNodes
looper.run(checkNodesConnected(allNodes))
# Node B sends multiple declarations of node D's 0th protocol instance as
# primary to all nodes
for i in range(5):
# B.send(Primary(D.name, 0, B.viewNo))
B.send(primaryByNode(D.name, B, 0))
# No node from node A, node C, node D(node B is malicious anyway so not
# considering it) should have more than one primary declaration for node
# D since node D is slow. The one primary declaration for node D,
# that nodes A, C and D might have would be because of node B
def x():
primDecs = [p[0] for p in node.elector.primaryDeclarations[0].values()]
assert primDecs.count(D.name) <= 1
# also have to take into account the catchup procedure
timeout = waits.expectedPoolNominationTimeout(len(allNodes)) + \
waits.expectedPoolCatchupTime(len(allNodes))
for node in (A, C, D):
looper.run(eventually(x, retryWait=.5, timeout=timeout))
timeout = waits.expectedPoolElectionTimeout(
len(allNodes)) + delaySelfNomination
ensureElectionsDone(looper=looper, nodes=allNodes, customTimeout=timeout)
# Node D should not have any primary replica
assert not D.hasPrimary
def test_view_change_on_start(tconf, txnPoolNodeSet, looper, sdk_pool_handle,
sdk_wallet_client):
"""
Do view change on a without any requests
"""
old_view_no = txnPoolNodeSet[0].viewNo
master_primary = get_master_primary_node(txnPoolNodeSet)
other_nodes = [n for n in txnPoolNodeSet if n != master_primary]
delay_3pc = 10
delay_3pc_messages(txnPoolNodeSet, 0, delay_3pc)
sent_batches = 2
sdk_send_random_requests(looper, sdk_pool_handle, sdk_wallet_client,
sent_batches * tconf.Max3PCBatchSize)
def chk1():
t_root, s_root = check_uncommitteds_equal(other_nodes)
assert master_primary.domainLedger.uncommittedRootHash != t_root
assert master_primary.states[DOMAIN_LEDGER_ID].headHash != s_root
looper.run(eventually(chk1, retryWait=1))
timeout = tconf.PerfCheckFreq + \
waits.expectedPoolElectionTimeout(len(txnPoolNodeSet))
waitForViewChange(looper,
txnPoolNodeSet,
old_view_no + 1,
customTimeout=timeout)
ensure_all_nodes_have_same_data(looper, nodes=txnPoolNodeSet)
check_uncommitteds_equal(txnPoolNodeSet)
reset_delays_and_process_delayeds(txnPoolNodeSet)
sdk_send_random_and_check(looper,
txnPoolNodeSet,
sdk_pool_handle,
sdk_wallet_client,
2 * Max3PCBatchSize,
add_delay_to_timeout=delay_3pc)
ensure_all_nodes_have_same_data(looper, nodes=txnPoolNodeSet)
def checkProtocolInstanceSetup(looper: Looper,
nodes: Sequence[TestNode],
retryWait: float = 1,
customTimeout: float = None,
instances: Sequence[int] = None):
timeout = customTimeout or waits.expectedPoolElectionTimeout(len(nodes))
checkEveryProtocolInstanceHasOnlyOnePrimary(looper=looper,
nodes=nodes,
retryWait=retryWait,
timeout=timeout,
instances_list=instances)
checkEveryNodeHasAtMostOnePrimary(looper=looper,
nodes=nodes,
retryWait=retryWait,
customTimeout=timeout)
primaryReplicas = {replica.instId: replica
for node in nodes
for replica in node.replicas.values() if replica.isPrimary}
return [r[1] for r in
sorted(primaryReplicas.items(), key=operator.itemgetter(0))]
def testPrimaryElectionCase5(case5Setup, looper, txnPoolNodeSet):
"""
Case 5 - A node making primary declarations for a multiple other nodes.
Consider 4 nodes A, B, C, and D. Lets say node B is malicious and
declares node C as primary to all nodes.
Again node B declares node D as primary to all nodes.
"""
A, B, C, D = txnPoolNodeSet
looper.run(checkNodesConnected(txnPoolNodeSet))
BRep = Replica.generateName(B.name, 0)
CRep = Replica.generateName(C.name, 0)
DRep = Replica.generateName(D.name, 0)
# Node B first sends PRIMARY msgs for Node C to all nodes
# B.send(Primary(CRep, 0, B.viewNo))
B.send(primaryByNode(CRep, B, 0))
# Node B sends PRIMARY msgs for Node D to all nodes
# B.send(Primary(DRep, 0, B.viewNo))
B.send(primaryByNode(DRep, B, 0))
# Ensure elections are done
# also have to take into account the catchup procedure
timeout = waits.expectedPoolElectionTimeout(len(txnPoolNodeSet)) + \
waits.expectedPoolCatchupTime(len(txnPoolNodeSet)) + \
delayOfElectionDone
ensureElectionsDone(looper=looper, nodes=txnPoolNodeSet, customTimeout=timeout)
# All nodes from node A, node C, node D(node B is malicious anyway so not
# considering it) should have primary declarations for node C from node B
# since node B first nominated node C
for node in [A, C, D]:
logger.debug(
"node {} should have primary declaration for C from node B"
.format(node))
assert node.elector.primaryDeclarations[0][BRep][0] == CRep
def test_restart_half_to_lower_view(looper, txnPoolNodeSet, tconf, tdir, allPluginsPath,
sdk_pool_handle, sdk_wallet_client):
# Add transaction to ledger
sdk_send_random_and_check(looper, txnPoolNodeSet, sdk_pool_handle, sdk_wallet_client, 1)
# Move to higher view
ensure_view_change_complete(looper, txnPoolNodeSet)
# Restart half of nodes
tm = tconf.ToleratePrimaryDisconnection + waits.expectedPoolElectionTimeout(len(txnPoolNodeSet))
nodes_before_restart = txnPoolNodeSet.copy()
restart_nodes(looper, txnPoolNodeSet, txnPoolNodeSet[2:], tconf, tdir, allPluginsPath,
after_restart_timeout=tm, start_one_by_one=False)
# Check that nodes didn't think they may have inconsistent 3PC state
for node in nodes_before_restart:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 0
# Check that nodes don't think they may have inconsistent 3PC state
for node in txnPoolNodeSet:
assert node.spylog.count(node.on_inconsistent_3pc_state) == 0
# Check that all nodes are still functional
sdk_ensure_pool_functional(looper, txnPoolNodeSet, sdk_wallet_client, sdk_pool_handle, num_reqs=2, num_batches=1)
def test_quorum_after_f_plus_2_nodes_including_primary_turned_off_and_later_on(
looper, allPluginsPath, tdir, tconf, txnPoolNodeSet, sdk_pool_handle,
sdk_wallet_client):
timeout = sdk_eval_timeout(1, len(txnPoolNodeSet))
nodes = txnPoolNodeSet
sdk_send_random_and_check(looper, txnPoolNodeSet, sdk_pool_handle,
sdk_wallet_client, 1)
stop_node(nodes[0], looper, nodes)
waitForViewChange(looper, nodes[1:], expectedViewNo=1)
ensureElectionsDone(looper,
nodes[1:],
numInstances=getRequiredInstances(nodeCount))
sdk_send_random_and_check(looper, txnPoolNodeSet, sdk_pool_handle,
sdk_wallet_client, 1)
stop_node(nodes[1], looper, nodes)
looper.runFor(tconf.ToleratePrimaryDisconnection +
waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[2:], expectedViewNo=1)
sdk_reqs3 = sdk_send_random_requests(looper, sdk_pool_handle,
sdk_wallet_client, 1)
with pytest.raises(TimeoutError):
req_res = sdk_get_replies(looper, sdk_reqs3, timeout=timeout)
sdk_check_reply(req_res[0])
stop_node(nodes[2], looper, nodes)
looper.runFor(tconf.ToleratePrimaryDisconnection +
waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[3:], expectedViewNo=1)
sdk_reqs4 = sdk_send_random_requests(looper, sdk_pool_handle,
sdk_wallet_client, 1)
with pytest.raises(TimeoutError):
req_res = sdk_get_replies(looper, sdk_reqs4, timeout=timeout)
sdk_check_reply(req_res[0])
nodes[2] = start_stopped_node(nodes[2], looper, tconf, tdir,
allPluginsPath)
looper.runFor(waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[3:], expectedViewNo=1)
sdk_reqs5 = sdk_send_random_requests(looper, sdk_pool_handle,
sdk_wallet_client, 1)
with pytest.raises(TimeoutError):
req_res = sdk_get_replies(looper, sdk_reqs5, timeout=timeout)
sdk_check_reply(req_res[0])
nodes[1] = start_stopped_node(nodes[1], looper, tconf, tdir,
allPluginsPath)
ensureElectionsDone(looper,
nodes[1:],
numInstances=getRequiredInstances(nodeCount))
checkViewNoForNodes(nodes[1:], expectedViewNo=1)
sdk_send_random_and_check(looper, txnPoolNodeSet, sdk_pool_handle,
sdk_wallet_client, 1)
nodes[0] = start_stopped_node(nodes[0], looper, tconf, tdir,
allPluginsPath)
ensureElectionsDone(looper,
nodes,
numInstances=getRequiredInstances(nodeCount))
checkViewNoForNodes(nodes, expectedViewNo=1)
sdk_send_random_and_check(looper, txnPoolNodeSet, sdk_pool_handle,
sdk_wallet_client, 1)
def test_node_erases_last_sent_pp_key_on_propagate_primary_after_pool_restart(
looper, txnPoolNodeSet, sdk_pool_handle, sdk_wallet_client,
tconf, tdir, allPluginsPath, chkFreqPatched):
# Get a node with a backup primary replica and the rest of the nodes
replica = getPrimaryReplica(txnPoolNodeSet, instId=backup_inst_id)
node = replica.node
other_nodes = set(txnPoolNodeSet) - {node}
# Send some 3PC-batches and wait until the replica orders the 3PC-batches
sdk_send_batches_of_random(looper, txnPoolNodeSet,
sdk_pool_handle, sdk_wallet_client,
num_reqs=7, num_batches=7,
timeout=tconf.Max3PCBatchWait)
looper.run(
eventually(lambda: assertExp(replica.last_ordered_3pc == (0, 7)),
retryWait=1,
timeout=waits.expectedTransactionExecutionTime(nodeCount)))
# Check view no of the node and lastPrePrepareSeqNo of the replica
assert node.viewNo == 0
assert replica.lastPrePrepareSeqNo == 7
assert replica.h == 6
assert replica.H == 6 + LOG_SIZE
# Ensure that there is a stored last sent PrePrepare key on the node
assert LAST_SENT_PRE_PREPARE in node.nodeStatusDB
# Stop the node containing the replica
disconnect_node_and_ensure_disconnected(looper,
txnPoolNodeSet,
node.name,
stopNode=True)
looper.removeProdable(node)
txnPoolNodeSet.remove(node)
# Restart the rest of the nodes and wait for primary elections done
for n in other_nodes:
disconnect_node_and_ensure_disconnected(looper,
txnPoolNodeSet,
n.name,
timeout=nodeCount,
stopNode=True)
looper.removeProdable(n)
txnPoolNodeSet.remove(n)
for n in other_nodes:
txnPoolNodeSet.append(start_stopped_node(n, looper, tconf, tdir, allPluginsPath))
looper.run(checkNodesConnected(txnPoolNodeSet,
customTimeout=waits.expectedPoolInterconnectionTime(nodeCount)))
def chk():
for n in txnPoolNodeSet:
for r in n.replicas.values():
assert r.hasPrimary
looper.run(eventually(chk, retryWait=1, timeout=waits.expectedPoolElectionTimeout(nodeCount)))
# Start the stopped node
node = start_stopped_node(node, looper, tconf, tdir, allPluginsPath)
txnPoolNodeSet.append(node)
looper.run(checkNodesConnected(txnPoolNodeSet))
ensureElectionsDone(looper, txnPoolNodeSet)
replica = node.replicas[backup_inst_id]
# Verify that the node has erased the stored last sent PrePrepare key
assert LAST_SENT_PRE_PREPARE not in node.nodeStatusDB
# Verify correspondingly that after the propagate primary procedure
# the replica (which must again be the primary in its instance) has
# not restored lastPrePrepareSeqNo, not adjusted last_ordered_3pc and
# not shifted the watermarks
assert node.viewNo == 0
assert replica.isPrimary
assert replica.lastPrePrepareSeqNo == 0
assert replica.last_ordered_3pc == (0, 0)
assert replica.h == 0
assert replica.H == 0 + LOG_SIZE
# Send a 3PC-batch and ensure that the replica orders it
sdk_send_batches_of_random(looper, txnPoolNodeSet,
sdk_pool_handle, sdk_wallet_client,
num_reqs=1, num_batches=1,
timeout=tconf.Max3PCBatchWait)
looper.run(
eventually(lambda: assertExp(replica.last_ordered_3pc == (0, 1)),
retryWait=1,
timeout=waits.expectedTransactionExecutionTime(nodeCount)))
def test_quorum_after_f_plus_2_nodes_including_primary_turned_off_and_later_on(
looper, allPluginsPath, tdir, tconf,
txnPoolNodeSet,
sdk_pool_handle,
sdk_wallet_client):
timeout = sdk_eval_timeout(1, len(txnPoolNodeSet))
nodes = txnPoolNodeSet
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle,
sdk_wallet_client,
1)
stop_node(nodes[0], looper, nodes)
waitForViewChange(looper, nodes[1:], expectedViewNo=1)
ensureElectionsDone(looper, nodes[1:],
instances_list=range(getRequiredInstances(nodeCount)))
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle,
sdk_wallet_client,
1)
stop_node(nodes[1], looper, nodes)
looper.runFor(tconf.ToleratePrimaryDisconnection +
waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[2:], expectedViewNo=1)
sdk_reqs3 = sdk_send_random_requests(looper,
sdk_pool_handle,
sdk_wallet_client,
1)
with pytest.raises(PoolLedgerTimeoutException):
req_res = sdk_get_replies(looper, sdk_reqs3, timeout=timeout)
sdk_check_reply(req_res[0])
stop_node(nodes[2], looper, nodes)
looper.runFor(tconf.ToleratePrimaryDisconnection +
waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[3:], expectedViewNo=1)
sdk_reqs4 = sdk_send_random_requests(looper,
sdk_pool_handle,
sdk_wallet_client,
1)
with pytest.raises(PoolLedgerTimeoutException):
req_res = sdk_get_replies(looper, sdk_reqs4, timeout=timeout)
sdk_check_reply(req_res[0])
nodes[2] = start_stopped_node(nodes[2], looper, tconf, tdir, allPluginsPath)
looper.runFor(waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[3:], expectedViewNo=1)
sdk_reqs5 = sdk_send_random_requests(looper,
sdk_pool_handle,
sdk_wallet_client,
1)
with pytest.raises(PoolLedgerTimeoutException):
req_res = sdk_get_replies(looper, sdk_reqs5, timeout=timeout)
sdk_check_reply(req_res[0])
nodes[1] = start_stopped_node(nodes[1], looper, tconf, tdir, allPluginsPath)
ensureElectionsDone(looper, nodes[1:],
instances_list=range(getRequiredInstances(nodeCount)),
customTimeout=60)
checkViewNoForNodes(nodes[1:], expectedViewNo=1)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle,
sdk_wallet_client,
1)
nodes[0] = start_stopped_node(nodes[0], looper, tconf, tdir, allPluginsPath)
ensureElectionsDone(looper, nodes,
instances_list=range(getRequiredInstances(nodeCount)),
customTimeout=60)
checkViewNoForNodes(nodes, expectedViewNo=1)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle,
sdk_wallet_client,
1)
def test_quorum_after_f_plus_2_nodes_but_not_primary_turned_off_and_later_on(
looper, allPluginsPath, tdir, tconf,
txnPoolNodeSet, sdk_pool_handle, sdk_wallet_client):
nodes = txnPoolNodeSet
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle,
sdk_wallet_client,
1)
stop_node(nodes[4], looper, nodes)
looper.runFor(tconf.ToleratePrimaryDisconnection +
waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[:4], expectedViewNo=0)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle,
sdk_wallet_client,
1)
stop_node(nodes[3], looper, nodes)
looper.runFor(tconf.ToleratePrimaryDisconnection +
waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[:3], expectedViewNo=0)
sdk_reqs3 = sdk_send_random_requests(looper,
sdk_pool_handle,
sdk_wallet_client,
1)
with pytest.raises(PoolLedgerTimeoutException):
req_res = sdk_get_replies(looper, sdk_reqs3)
sdk_check_reply(req_res[0])
stop_node(nodes[2], looper, nodes)
looper.runFor(tconf.ToleratePrimaryDisconnection +
waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[:2], expectedViewNo=0)
sdk_reqs4 = sdk_send_random_requests(looper,
sdk_pool_handle,
sdk_wallet_client,
1)
with pytest.raises(PoolLedgerTimeoutException):
req_res = sdk_get_replies(looper, sdk_reqs4)
sdk_check_reply(req_res[0])
nodes[4] = start_stopped_node(nodes[4], looper, tconf, tdir, allPluginsPath)
looper.runFor(waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[:2] + nodes[4:], expectedViewNo=0)
sdk_reqs5 = sdk_send_random_requests(looper,
sdk_pool_handle,
sdk_wallet_client,
1)
with pytest.raises(PoolLedgerTimeoutException):
req_res = sdk_get_replies(looper, sdk_reqs5)
sdk_check_reply(req_res[0])
nodes[3] = start_stopped_node(nodes[3], looper, tconf, tdir, allPluginsPath)
ensureElectionsDone(looper, nodes[:2] + nodes[3:],
instances_list=range(getRequiredInstances(nodeCount)))
checkViewNoForNodes(nodes[:2] + nodes[3:], expectedViewNo=0)
sdk_reqs6 = sdk_send_random_requests(looper,
sdk_pool_handle,
sdk_wallet_client,
1)
sdk_get_replies(looper, sdk_reqs6)
nodes[2] = start_stopped_node(nodes[2], looper, tconf, tdir, allPluginsPath)
ensureElectionsDone(looper, nodes,
instances_list=range(getRequiredInstances(nodeCount)))
checkViewNoForNodes(nodes, expectedViewNo=0)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle,
sdk_wallet_client,
1)
def test_quorum_after_f_plus_2_nodes_but_not_primary_turned_off_and_later_on(
looper, allPluginsPath, tdir, tconf,
txnPoolNodeSet, sdk_pool_handle, sdk_wallet_client):
nodes = txnPoolNodeSet
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle,
sdk_wallet_client,
1)
stop_node(nodes[4], looper, nodes)
looper.runFor(tconf.ToleratePrimaryDisconnection +
waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[:4], expectedViewNo=0)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle,
sdk_wallet_client,
1)
stop_node(nodes[3], looper, nodes)
looper.runFor(tconf.ToleratePrimaryDisconnection +
waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[:3], expectedViewNo=0)
sdk_reqs3 = sdk_send_random_requests(looper,
sdk_pool_handle,
sdk_wallet_client,
1)
with pytest.raises(PoolLedgerTimeoutException):
req_res = sdk_get_replies(looper, sdk_reqs3)
sdk_check_reply(req_res[0])
stop_node(nodes[2], looper, nodes)
looper.runFor(tconf.ToleratePrimaryDisconnection +
waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[:2], expectedViewNo=0)
sdk_reqs4 = sdk_send_random_requests(looper,
sdk_pool_handle,
sdk_wallet_client,
1)
with pytest.raises(PoolLedgerTimeoutException):
req_res = sdk_get_replies(looper, sdk_reqs4)
sdk_check_reply(req_res[0])
nodes[4] = start_stopped_node(nodes[4], looper, tconf, tdir, allPluginsPath)
looper.runFor(waits.expectedPoolElectionTimeout(len(nodes)))
checkViewNoForNodes(nodes[:2] + nodes[4:], expectedViewNo=0)
sdk_reqs5 = sdk_send_random_requests(looper,
sdk_pool_handle,
sdk_wallet_client,
1)
with pytest.raises(PoolLedgerTimeoutException):
req_res = sdk_get_replies(looper, sdk_reqs5)
sdk_check_reply(req_res[0])
nodes[3] = start_stopped_node(nodes[3], looper, tconf, tdir, allPluginsPath)
ensureElectionsDone(looper, nodes[:2] + nodes[3:],
numInstances=getRequiredInstances(nodeCount))
checkViewNoForNodes(nodes[:2] + nodes[3:], expectedViewNo=0)
sdk_reqs6 = sdk_send_random_requests(looper,
sdk_pool_handle,
sdk_wallet_client,
1)
sdk_get_replies(looper, [sdk_reqs3[0], sdk_reqs4[0], sdk_reqs5[0], sdk_reqs6[0]])
nodes[2] = start_stopped_node(nodes[2], looper, tconf, tdir, allPluginsPath)
ensureElectionsDone(looper, nodes,
numInstances=getRequiredInstances(nodeCount))
checkViewNoForNodes(nodes, expectedViewNo=0)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle,
sdk_wallet_client,
1)
def test_disconnected_node_with_lagged_view_pulls_up_its_view_on_reconnection(
looper, txnPoolNodeSet, sdk_wallet_client, sdk_pool_handle):
"""
Verifies that a disconnected node with a lagged view accepts
the current view from the other nodes on re-connection.
Steps:
1. Provoke view change to 1.
2. Ensure that all the nodes complete view change to 1.
3. Disconnect one node from the rest of the nodes in the pool.
4. Provoke view change to 2.
5. Ensure that that all the nodes except for the disconnected one complete
view change to 2 and the disconnected node remains in the view 1.
6. Provoke view change to 3.
5. Ensure that that all the nodes except for the disconnected one complete
view change to 3 and the disconnected node remains in the view 1.
8. Connect the disconnected node to the rest of the nodes in the pool.
9. Ensure that the re-connected node completes view change to 3.
10. Ensure that all the nodes participate in consensus.
"""
checkViewNoForNodes(txnPoolNodeSet, 0)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle, sdk_wallet_client, 1)
ensure_view_change(looper, txnPoolNodeSet)
ensureElectionsDone(looper, txnPoolNodeSet)
ensure_all_nodes_have_same_data(looper, txnPoolNodeSet)
checkViewNoForNodes(txnPoolNodeSet, 1)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle, sdk_wallet_client, 1)
lagged_node = getNonPrimaryReplicas(txnPoolNodeSet)[-1].node
disconnect_node_and_ensure_disconnected(looper,
txnPoolNodeSet,
lagged_node,
stopNode=False)
other_nodes = list(set(txnPoolNodeSet) - {lagged_node})
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle, sdk_wallet_client, 1)
ensure_view_change(looper, other_nodes)
ensureElectionsDone(looper, other_nodes,
instances_list=range(getRequiredInstances(len(txnPoolNodeSet))))
ensure_all_nodes_have_same_data(looper, other_nodes)
checkViewNoForNodes(other_nodes, 2)
checkViewNoForNodes([lagged_node], 1)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle, sdk_wallet_client, 1)
ensure_view_change(looper, other_nodes)
ensureElectionsDone(looper, other_nodes,
instances_list=range(getRequiredInstances(len(txnPoolNodeSet))))
ensure_all_nodes_have_same_data(looper, other_nodes)
checkViewNoForNodes(other_nodes, 3)
checkViewNoForNodes([lagged_node], 1)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle, sdk_wallet_client, 1)
reconnect_node_and_ensure_connected(looper, txnPoolNodeSet, lagged_node)
waitForViewChange(looper, [lagged_node], 3,
customTimeout=waits.expectedPoolElectionTimeout(
len(txnPoolNodeSet)))
ensureElectionsDone(looper, txnPoolNodeSet)
ensure_all_nodes_have_same_data(looper, txnPoolNodeSet)
checkViewNoForNodes(txnPoolNodeSet, 3)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle, sdk_wallet_client, 1)
ensure_all_nodes_have_same_data(looper, txnPoolNodeSet)
def test_disconnected_node_with_lagged_view_pulls_up_its_view_on_reconnection(
looper, txnPoolNodeSet, sdk_wallet_client, sdk_pool_handle):
"""
Verifies that a disconnected node with a lagged view accepts
the current view from the other nodes on re-connection.
Steps:
1. Provoke view change to 1.
2. Ensure that all the nodes complete view change to 1.
3. Disconnect one node from the rest of the nodes in the pool.
4. Provoke view change to 2.
5. Ensure that that all the nodes except for the disconnected one complete
view change to 2 and the disconnected node remains in the view 1.
6. Provoke view change to 3.
5. Ensure that that all the nodes except for the disconnected one complete
view change to 3 and the disconnected node remains in the view 1.
8. Connect the disconnected node to the rest of the nodes in the pool.
9. Ensure that the re-connected node completes view change to 3.
10. Ensure that all the nodes participate in consensus.
"""
checkViewNoForNodes(txnPoolNodeSet, 0)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle, sdk_wallet_client, 1)
ensure_view_change(looper, txnPoolNodeSet)
ensureElectionsDone(looper, txnPoolNodeSet)
ensure_all_nodes_have_same_data(looper, txnPoolNodeSet)
checkViewNoForNodes(txnPoolNodeSet, 1)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle, sdk_wallet_client, 1)
lagged_node = getNonPrimaryReplicas(txnPoolNodeSet)[-1].node
disconnect_node_and_ensure_disconnected(looper,
txnPoolNodeSet,
lagged_node,
stopNode=False)
other_nodes = list(set(txnPoolNodeSet) - {lagged_node})
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle, sdk_wallet_client, 1)
ensure_view_change(looper, other_nodes)
ensureElectionsDone(looper, other_nodes,
numInstances=getRequiredInstances(len(txnPoolNodeSet)))
ensure_all_nodes_have_same_data(looper, other_nodes)
checkViewNoForNodes(other_nodes, 2)
checkViewNoForNodes([lagged_node], 1)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle, sdk_wallet_client, 1)
ensure_view_change(looper, other_nodes)
ensureElectionsDone(looper, other_nodes,
numInstances=getRequiredInstances(len(txnPoolNodeSet)))
ensure_all_nodes_have_same_data(looper, other_nodes)
checkViewNoForNodes(other_nodes, 3)
checkViewNoForNodes([lagged_node], 1)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle, sdk_wallet_client, 1)
reconnect_node_and_ensure_connected(looper, txnPoolNodeSet, lagged_node)
waitForViewChange(looper, [lagged_node], 3,
customTimeout=waits.expectedPoolElectionTimeout(
len(txnPoolNodeSet)))
ensureElectionsDone(looper, txnPoolNodeSet)
ensure_all_nodes_have_same_data(looper, txnPoolNodeSet)
checkViewNoForNodes(txnPoolNodeSet, 3)
sdk_send_random_and_check(looper, txnPoolNodeSet,
sdk_pool_handle, sdk_wallet_client, 1)
ensure_all_nodes_have_same_data(looper, txnPoolNodeSet)
