[SDFAB-642][SDFAB-643] Fix device disconnection when an instance rejoins
Changes in the GDP are done to make sure the default availability is always
updated through the two step process defined in the provider and to prevent
the disconnection of the devices when an instance rejoins the cluster
Finally, the patch improves logging inside the GossipDeviceStore and improve
the stability of PiPipeconfWatchdogManager by getting rid of the timer and
timer task and by executing the event listeners inside an executor.
Change-Id: Ibc6ce711e15e86bde05dbf3b1c37d2a93516fae3
diff --git a/core/store/dist/src/main/java/org/onosproject/store/device/impl/GossipDeviceStore.java b/core/store/dist/src/main/java/org/onosproject/store/device/impl/GossipDeviceStore.java
index 5760e23..2bd4dfe 100644
--- a/core/store/dist/src/main/java/org/onosproject/store/device/impl/GossipDeviceStore.java
+++ b/core/store/dist/src/main/java/org/onosproject/store/device/impl/GossipDeviceStore.java
@@ -1037,6 +1037,20 @@
}
@Override
+ public DeviceDescription getDeviceDescription(ProviderId providerId, DeviceId deviceId) {
+ if (devices.containsKey(deviceId)) {
+ Map<ProviderId, DeviceDescriptions> descs = getOrCreateDeviceDescriptionsMap(deviceId);
+ synchronized (descs) {
+ DeviceDescriptions deviceDescriptions = descs.get(providerId);
+ return deviceDescriptions != null ? deviceDescriptions.getDeviceDesc().value() : null;
+ }
+ } else {
+ log.warn("Device {} does not exist in store", deviceId);
+ }
+ return null;
+ }
+
+ @Override
public boolean isAvailable(DeviceId deviceId) {
return availableDevices.contains(deviceId);
}
@@ -1403,7 +1417,13 @@
* @param advertisement to respond to
*/
private void handleAdvertisement(DeviceAntiEntropyAdvertisement advertisement) {
-
+ /*
+ * NOTE that when an instance rejoins the cluster, it will generate
+ * device events and send to the local apps through the delegate. This
+ * approach might be not the best if the apps are not enough robust or
+ * if there is no proper coordination in the cluster. Also, note that
+ * any ECMap will act on the same way during the bootstrap process
+ */
final NodeId sender = advertisement.sender();
Map<DeviceFragmentId, Timestamp> devAds = new HashMap<>(advertisement.deviceFingerPrints());
@@ -1437,9 +1457,12 @@
if (advDevTimestamp == null || lProvDevice.isNewerThan(
advDevTimestamp)) {
// remote does not have it or outdated, suggest
+ log.trace("send to {} device update {} for {}", sender, lProvDevice, deviceId);
notifyPeer(sender, new InternalDeviceEvent(provId, deviceId, lProvDevice));
} else if (!lProvDevice.timestamp().equals(advDevTimestamp)) {
// local is outdated, request
+ log.trace("need update {} < {} for device {} from {}", lProvDevice.timestamp(),
+ advDevTimestamp, deviceId, sender);
reqDevices.add(devFragId);
}
@@ -1456,10 +1479,12 @@
if (advPortTimestamp == null || lPort.isNewerThan(
advPortTimestamp)) {
// remote does not have it or outdated, suggest
+ log.trace("send to {} port update {} for {}/{}", sender, lPort, deviceId, num);
notifyPeer(sender, new InternalPortStatusEvent(provId, deviceId, lPort));
} else if (!lPort.timestamp().equals(advPortTimestamp)) {
// local is outdated, request
- log.trace("need update {} < {}", lPort.timestamp(), advPortTimestamp);
+ log.trace("need update {} < {} for port {} from {}", lPort.timestamp(),
+ advPortTimestamp, num, sender);
reqPorts.add(portFragId);
}
@@ -1483,12 +1508,15 @@
if (localLatest == null || (rOffline != null && rOffline.compareTo(localLatest) > 0)) {
// remote offline timestamp suggests that the
// device is off-line
+ log.trace("remote offline timestamp from {} suggests that the device {} is off-line",
+ sender, deviceId);
markOfflineInternal(deviceId, rOffline);
}
Timestamp lOffline = offline.get(deviceId);
if (lOffline != null && rOffline == null) {
// locally offline, but remote is online, suggest offline
+ log.trace("suggest to {} sthat the device {} is off-line", sender, deviceId);
notifyPeer(sender, new InternalDeviceStatusChangeEvent(deviceId, lOffline, false));
}