Forget links if QoS verbs fail to arrive

This commit is contained in:
Joseph Henry 2022-09-07 09:08:13 -07:00
parent 0e23ba8f1a
commit 1d1843bf3b
No known key found for this signature in database
GPG key ID: C45B33FF5EBC9344
2 changed files with 19 additions and 11 deletions

View file

@ -493,7 +493,8 @@ void Bond::receivedQoS(const SharedPtr<Path>& path, int64_t now, int count, uint
if (pathIdx == ZT_MAX_PEER_NETWORK_PATHS) { if (pathIdx == ZT_MAX_PEER_NETWORK_PATHS) {
return; return;
} }
// debug("received QoS packet (sampling %d frames) via %s", count, pathToStr(path).c_str()); _paths[pathIdx].lastQoSReceived = now;
debug("received QoS packet (sampling %d frames) via %s", count, pathToStr(path).c_str());
// Look up egress times and compute latency values for each record // Look up egress times and compute latency values for each record
std::map<uint64_t, uint64_t>::iterator it; std::map<uint64_t, uint64_t>::iterator it;
for (int j = 0; j < count; j++) { for (int j = 0; j < count; j++) {
@ -799,7 +800,7 @@ void Bond::sendQOS_MEASUREMENT(void* tPtr, int pathIdx, int64_t localSocket, con
Packet outp(_peer->_id.address(), RR->identity.address(), Packet::VERB_QOS_MEASUREMENT); Packet outp(_peer->_id.address(), RR->identity.address(), Packet::VERB_QOS_MEASUREMENT);
char qosData[ZT_QOS_MAX_PACKET_SIZE]; char qosData[ZT_QOS_MAX_PACKET_SIZE];
int16_t len = generateQoSPacket(pathIdx, _now, qosData); int16_t len = generateQoSPacket(pathIdx, _now, qosData);
// debug("sending QOS via link %s (len=%d)", pathToStr(_paths[pathIdx].p).c_str(), len); debug("sending QOS via link %s (len=%d)", pathToStr(_paths[pathIdx].p).c_str(), len);
if (len) { if (len) {
outp.append(qosData, len); outp.append(qosData, len);
if (atAddress) { if (atAddress) {
@ -895,6 +896,9 @@ void Bond::curateBond(int64_t now, bool rebuildBond)
continue; continue;
} }
// Whether this path is still in its trial period
bool inTrial = (now - _paths[i].whenNominated) < ZT_BOND_OPTIMIZE_INTERVAL;
/** /**
* Remove expired or invalid links from bond * Remove expired or invalid links from bond
*/ */
@ -905,7 +909,7 @@ void Bond::curateBond(int64_t now, bool rebuildBond)
_paths[i].p = SharedPtr<Path>(); _paths[i].p = SharedPtr<Path>();
continue; continue;
} }
if ((now - _paths[i].p->_lastIn) > (ZT_PEER_EXPIRED_PATH_TRIAL_PERIOD)) { if ((now - _paths[i].lastEligibility) > (ZT_PEER_EXPIRED_PATH_TRIAL_PERIOD) && ! inTrial) {
log("link (%s) has expired or is invalid, removing from bond", pathToStr(_paths[i].p).c_str()); log("link (%s) has expired or is invalid, removing from bond", pathToStr(_paths[i].p).c_str());
_paths[i] = NominatedPath(); _paths[i] = NominatedPath();
_paths[i].p = SharedPtr<Path>(); _paths[i].p = SharedPtr<Path>();
@ -930,14 +934,13 @@ void Bond::curateBond(int64_t now, bool rebuildBond)
bool acceptableAge = _isLeaf ? (_paths[i].p->age(now) < (_failoverInterval + _downDelay)) : _paths[i].alive; bool acceptableAge = _isLeaf ? (_paths[i].p->age(now) < (_failoverInterval + _downDelay)) : _paths[i].alive;
// Whether we've waited long enough since the link last came online // Whether we've waited long enough since the link last came online
bool satisfiedUpDelay = (now - _paths[i].lastAliveToggle) >= _upDelay; bool satisfiedUpDelay = (now - _paths[i].lastAliveToggle) >= _upDelay;
// Whether this path is still in its trial period // How long since the last QoS was received (Must be less than ZT_PEER_PATH_EXPIRATION since the remote peer's _qosSendInterval isn't known)
bool inTrial = (now - _paths[i].whenNominated) < ZT_BOND_OPTIMIZE_INTERVAL; bool acceptableQoSAge = _paths[i].lastQoSReceived == 0 || ((now - _paths[i].lastQoSReceived) < ZT_PEER_EXPIRED_PATH_TRIAL_PERIOD);
// if (includeRefractoryPeriod && _paths[i].refractoryPeriod) { currEligibility = _paths[i].allowed() && ((acceptableAge && satisfiedUpDelay && acceptableQoSAge) || inTrial);
// As long as the refractory period value has not fully drained this path is not eligible
// currEligibility = false; if (currEligibility) {
//} _paths[i].lastEligibility = now;
currEligibility = _paths[i].allowed() && ((acceptableAge && satisfiedUpDelay) || inTrial); }
// debug("[%d] allowed=%d, acceptableAge=%d, satisfiedUpDelay=%d, inTrial=%d ==== %d", i, _paths[i].allowed(), acceptableAge, satisfiedUpDelay, inTrial, currEligibility);
/** /**
* Note eligibility state change (if any) and take appropriate action * Note eligibility state change (if any) and take appropriate action
@ -949,6 +952,7 @@ void Bond::curateBond(int64_t now, bool rebuildBond)
if (currEligibility == 1) { if (currEligibility == 1) {
log("link %s is eligible", pathToStr(_paths[i].p).c_str()); log("link %s is eligible", pathToStr(_paths[i].p).c_str());
} }
debug("\t[%d] allowed=%d, age=%d, qa=%d, ud=%d, trial=%d", i, _paths[i].allowed(), acceptableAge, acceptableQoSAge, satisfiedUpDelay, inTrial);
dumpPathStatus(now, i); dumpPathStatus(now, i);
if (currEligibility) { if (currEligibility) {
rebuildBond = true; rebuildBond = true;

View file

@ -1235,6 +1235,7 @@ class Bond {
NominatedPath() NominatedPath()
: lastAckSent(0) : lastAckSent(0)
, lastAckReceived(0) , lastAckReceived(0)
, lastQoSReceived(0)
, unackedBytes(0) , unackedBytes(0)
, packetsReceivedSinceLastAck(0) , packetsReceivedSinceLastAck(0)
, lastQoSMeasurement(0) , lastQoSMeasurement(0)
@ -1243,6 +1244,7 @@ class Bond {
, lastAliveToggle(0) , lastAliveToggle(0)
, alive(false) , alive(false)
, eligible(true) , eligible(true)
, lastEligibility(0)
, whenNominated(0) , whenNominated(0)
, refractoryPeriod(0) , refractoryPeriod(0)
, ipvPref(0) , ipvPref(0)
@ -1359,6 +1361,7 @@ class Bond {
uint64_t lastAckSent; uint64_t lastAckSent;
uint64_t lastAckReceived; uint64_t lastAckReceived;
uint64_t lastQoSReceived;
uint64_t unackedBytes; uint64_t unackedBytes;
uint64_t packetsReceivedSinceLastAck; uint64_t packetsReceivedSinceLastAck;
@ -1368,6 +1371,7 @@ class Bond {
uint64_t lastAliveToggle; // The last time that the path was marked as "alive". uint64_t lastAliveToggle; // The last time that the path was marked as "alive".
bool alive; bool alive;
bool eligible; // State of eligibility at last check. Used for determining state changes. bool eligible; // State of eligibility at last check. Used for determining state changes.
uint64_t lastEligibility; // The last time that this path was eligible
uint64_t whenNominated; // Timestamp indicating when this path's trial period began. uint64_t whenNominated; // Timestamp indicating when this path's trial period began.
uint32_t refractoryPeriod; // Amount of time that this path will be prevented from becoming a member of a bond. uint32_t refractoryPeriod; // Amount of time that this path will be prevented from becoming a member of a bond.
uint8_t ipvPref; // IP version preference inherited from the physical link. uint8_t ipvPref; // IP version preference inherited from the physical link.