Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
grpclb: skip fallback if the LB is already in fallback mode (#8253)
Manually checks if the gRPCLB policy is already in fallback mode when trying to fallback due to receiving address update without LB addresses. 

Commit b956f88 added an invariant check in the FallbackModeTask runnable to ensure the task is fired only when the LB is not already in fallback mode. However, that commit missed the case that receiving address updates without LB addresses can trigger the run of FallbackModeTask runnable, because the existing implementation chose to reuse the code in FallbackModeTask. In such case, running FallbackModeTask could break the invariant check as the LB policy may already in fallback mode.

This change eliminates the reuse of FallbackModeTask for handling address update without LB address. That is, every time receiving address update, we manually check if it is already in fallback instead of reusing to FallbackModeTask perform the check.

Note there was a discussion brought up whether we should force entering fallback (shutdown existing subchannels) or we should still keep the balancer connection. Different languages have already diverged on this. Go shuts down the balancer connection and all subchannel connections to force using fallback addresses. C-core keep the balancer connection working and does not shutdown subchannels, only let fallback happens after the existing balancer connection and subchannel connections become broken. Java shuts down the balancer connection but not subchannels. This change does not try to change the existing behavior, but only fixes the invariant check breakage.

-------------------
See bug reported in  b/190700476
  • Loading branch information
voidzcy committed Jun 15, 2021
commit 64ee2574e02329acb28161931403fa7391e98cec
15 changes: 11 additions & 4 deletions grpclb/src/main/java/io/grpc/grpclb/GrpclbState.java
Original file line number Diff line number Diff line change
Expand Up @@ -259,11 +259,19 @@ void handleAddresses(
serviceName,
newLbAddressGroups,
newBackendServers);
fallbackBackendList = newBackendServers;
if (newLbAddressGroups.isEmpty()) {
// No balancer address: close existing balancer connection and enter fallback mode
// immediately.
// No balancer address: close existing balancer connection and prepare to enter fallback
// mode. If there is no successful backend connection, it enters fallback mode immediately.
// Otherwise, fallback does not happen until backend connections are lost. This behavior
// might be different from other languages (e.g., existing balancer connection is not
// closed in C-core), but we aren't changing it at this time.
shutdownLbComm();
syncContext.execute(new FallbackModeTask(NO_LB_ADDRESS_PROVIDED_STATUS));
if (!usingFallbackBackends) {
fallbackReason = NO_LB_ADDRESS_PROVIDED_STATUS;
cancelFallbackTimer();
maybeUseFallbackBackends();
}
} else {
startLbComm(newLbAddressGroups);
// Avoid creating a new RPC just because the addresses were updated, as it can cause a
Expand All @@ -281,7 +289,6 @@ void handleAddresses(
TimeUnit.MILLISECONDS, timerService);
}
}
fallbackBackendList = newBackendServers;
if (usingFallbackBackends) {
// Populate the new fallback backends to round-robin list.
useFallbackBackends();
Expand Down
29 changes: 27 additions & 2 deletions grpclb/src/test/java/io/grpc/grpclb/GrpclbLoadBalancerTest.java
Original file line number Diff line number Diff line change
Expand Up @@ -1452,8 +1452,11 @@ public void grpclbFallback_breakLbStreamBeforeFallbackTimerExpires() {
public void grpclbFallback_noBalancerAddress() {
InOrder inOrder = inOrder(helper, subchannelPool);

// Create just backend addresses
List<EquivalentAddressGroup> backendList = createResolvedBackendAddresses(2);
// Create 5 distinct backends
List<EquivalentAddressGroup> backends = createResolvedBackendAddresses(5);

// Name resolver gives the first two backend addresses
List<EquivalentAddressGroup> backendList = backends.subList(0, 2);
deliverResolvedAddresses(backendList, Collections.<EquivalentAddressGroup>emptyList());

assertThat(logs).containsAtLeast(
Expand All @@ -1474,6 +1477,28 @@ public void grpclbFallback_noBalancerAddress() {
.createOobChannel(ArgumentMatchers.<EquivalentAddressGroup>anyList(), anyString());
logs.clear();

/////////////////////////////////////////////////////////////////////////////////////////
// Name resolver sends new resolution results with new backend addr but no balancer addr
/////////////////////////////////////////////////////////////////////////////////////////
// Name resolver then gives the last three backends
backendList = backends.subList(2, 5);
deliverResolvedAddresses(backendList, Collections.<EquivalentAddressGroup>emptyList());

assertThat(logs).containsAtLeast(
"INFO: [grpclb-<api.google.com>] Using fallback backends",
"INFO: [grpclb-<api.google.com>] "
+ "Using RR list=[[[FakeSocketAddress-fake-address-2]/{}], "
+ "[[FakeSocketAddress-fake-address-3]/{}], "
+ "[[FakeSocketAddress-fake-address-4]/{}]], drop=[null, null, null]",
"INFO: [grpclb-<api.google.com>] "
+ "Update balancing state to CONNECTING: picks=[BUFFER_ENTRY], "
+ "drops=[null, null, null]")
.inOrder();

// Shift to use updated backends
fallbackTestVerifyUseOfFallbackBackendLists(inOrder, backendList);
logs.clear();

///////////////////////////////////////////////////////////////////////////////////////
// Name resolver sends new resolution results without any backend addr or balancer addr
///////////////////////////////////////////////////////////////////////////////////////
Expand Down