[
https://issues.apache.org/jira/browse/IGNITE-20955?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Rahul Mohan updated IGNITE-20955:
---------------------------------
Language: Java
> Issue in Reconnect Throttling
> ------------------------------
>
> Key: IGNITE-20955
> URL: https://issues.apache.org/jira/browse/IGNITE-20955
> Project: Ignite
> Issue Type: Bug
> Components: thin client
> Affects Versions: 2.14, 2.15
> Reporter: Rahul Mohan
> Priority: Major
>
> Encountered an issue where the below Reconnect throttling exception is thrown
> even when there are no connection issues to Ignite server.
> {code:java}
> org.apache.ignite.client.ClientConnectionException: Reconnect is not allowed
> due to applied throttling
> at
> org.apache.ignite.internal.client.thin.ReliableChannel$ClientChannelHolder.getOrCreateChannel(ReliableChannel.java:959)
> ~[ignite-core-2.14.0.jar!/:2.14.0]
> at
> org.apache.ignite.internal.client.thin.ReliableChannel$ClientChannelHolder.getOrCreateChannel(ReliableChannel.java:942)
> ~[ignite-core-2.14.0.jar!/:2.14.0]
> at
> org.apache.ignite.internal.client.thin.ReliableChannel$ClientChannelHolder.access$200(ReliableChannel.java:891)
> ~[ignite-core-2.14.0.jar!/:2.14.0]
> at
> org.apache.ignite.internal.client.thin.ReliableChannel.applyOnDefaultChannel(ReliableChannel.java:795)
> ~[ignite-core-2.14.0.jar!/:2.14.0]
> at
> org.apache.ignite.internal.client.thin.ReliableChannel.applyOnNodeChannelWithFallback(ReliableChannel.java:848)
> ~[ignite-core-2.14.0.jar!/:2.14.0]
> at
> org.apache.ignite.internal.client.thin.ReliableChannel.affinityService(ReliableChannel.java:315)
> ~[ignite-core-2.14.0.jar!/:2.14.0]
> at
> org.apache.ignite.internal.client.thin.TcpClientCache.cacheSingleKeyOperation(TcpClientCache.java:1084)
> ~[ignite-core-2.14.0.jar!/:2.14.0]{code}
> Disabling reconnect throttling avoids this issue and cache accesses works as
> expected, which provides further confirmation that there are no underlying
> connection issues.
> On digging into the thin client code i noticed that,
> applyReconnectionThrottling() doesnt actually check the connection status of
> the Channel but just tracks the timestamps of the last n (where n is
> reconnectThrottlingRetries ) connection/operations .
> {code:java}
> org.apache.ignite.internal.client.thin.ReliableChannel.java
> private ClientChannel getOrCreateChannel(boolean ignoreThrottling)
> throws ClientConnectionException, ClientAuthenticationException,
> ClientProtocolError {
> if (ch == null && !close) {
> synchronized (this) {
> if (close)
> return null;
> if (ch != null)
> return ch;
> if (!ignoreThrottling && applyReconnectionThrottling())
> throw new ClientConnectionException("Reconnect is not
> allowed due to applied throttling"); {code}
>
> {code:java}
> org.apache.ignite.internal.client.thin.ReliableChannel.java
> private boolean applyReconnectionThrottling() {
> if (reconnectRetries == null)
> return false;
> long ts = System.currentTimeMillis();
> for (int i = 0; i < reconnectRetries.length; i++) {
> if (ts - reconnectRetries[i] >=
> chCfg.getReconnectThrottlingPeriod()){
> reconnectRetries[i] = ts;
> return false;
> }
> }
> return true;
> }{code}
> Assuming my understanding is correct , its possible that depending on the
> timing of cache access operations within reconnectThrottlingPeriod requests
> are mistakenly throttled even when there are no prior connection/timeout
> issues.
> Example scenario:
> Using default value of throttling period and retries and using mock
> timestamps for conciseness
> reconnectThrottlingPeriod = 30_000L;
> reconnectThrottlingRetries = 3;
> ts=1000
> reconnectRetries[0] = 1000 -> returns false ,no throttlling exception
> ts=1005
> reconnectRetries[1] = 1005 -> returns false,no throttlling exception
> ts=1020
> reconnectRetries[2] = 1020 -> returns false,no throttlling exception
> ts=1025 -> returns true , reconnect throttlling exception thrown and
> operation aborted
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
