Repository: incubator-ignite Updated Branches: refs/heads/ignite-752-2 a12775621 -> 55f0eb569
ignite-752-2: supported connection check frequency even for cases when failure timeout is ignored; performance optimizations Project: http://git-wip-us.apache.org/repos/asf/incubator-ignite/repo Commit: http://git-wip-us.apache.org/repos/asf/incubator-ignite/commit/55f0eb56 Tree: http://git-wip-us.apache.org/repos/asf/incubator-ignite/tree/55f0eb56 Diff: http://git-wip-us.apache.org/repos/asf/incubator-ignite/diff/55f0eb56 Branch: refs/heads/ignite-752-2 Commit: 55f0eb56967d2cc9bdf62c3fb665521a59ddaf33 Parents: a127756 Author: Denis Magda <[email protected]> Authored: Wed Jul 29 09:15:29 2015 +0300 Committer: Denis Magda <[email protected]> Committed: Wed Jul 29 09:15:29 2015 +0300 ---------------------------------------------------------------------- .../ignite/spi/discovery/tcp/ServerImpl.java | 42 +++++++++++--------- 1 file changed, 23 insertions(+), 19 deletions(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/incubator-ignite/blob/55f0eb56/modules/core/src/main/java/org/apache/ignite/spi/discovery/tcp/ServerImpl.java ---------------------------------------------------------------------- diff --git a/modules/core/src/main/java/org/apache/ignite/spi/discovery/tcp/ServerImpl.java b/modules/core/src/main/java/org/apache/ignite/spi/discovery/tcp/ServerImpl.java index 547347c..b2be5b3 100644 --- a/modules/core/src/main/java/org/apache/ignite/spi/discovery/tcp/ServerImpl.java +++ b/modules/core/src/main/java/org/apache/ignite/spi/discovery/tcp/ServerImpl.java @@ -1787,6 +1787,9 @@ class ServerImpl extends TcpDiscoveryImpl { /** Connection check frequency. */ private long connCheckFreq; + /** Connection check threshold. */ + private long connCheckThreshold; + /** */ protected RingMessageWorker() { @@ -1799,19 +1802,22 @@ class ServerImpl extends TcpDiscoveryImpl { * Initializes connection check frequency. Used only when failure detection timeout is enabled. */ private void initConnectionCheckFrequency() { - if (spi.failureDetectionTimeoutEnabled()) { - for (int i = 3; i > 0; i--) { - connCheckFreq = spi.failureDetectionTimeout() / i; - - if (connCheckFreq > 0) - break; - } + if (spi.failureDetectionTimeoutEnabled()) + connCheckThreshold = spi.failureDetectionTimeout(); + else + connCheckThreshold = Math.min(spi.getSocketTimeout(), spi.getHeartbeatFrequency()); - assert connCheckFreq > 0; + for (int i = 3; i > 0; i--) { + connCheckFreq = connCheckThreshold / i; - if (log.isDebugEnabled()) - log.debug("Connection check frequency is calculated: " + connCheckFreq); + if (connCheckFreq > 10) + break; } + + assert connCheckFreq > 0; + + if (log.isDebugEnabled()) + log.debug("Connection check frequency is calculated: " + connCheckFreq); } /** @@ -2306,9 +2312,9 @@ class ServerImpl extends TcpDiscoveryImpl { // If node existed on connection initialization we should check // whether it has not gone yet. - if (nextNodeExists && pingNode(next)) - U.error(log, "Failed to send message to next node [msg=" + msg + - ", next=" + next + ']', err); + if (nextNodeExists) + log.warning("Failed to send message to next node [msg=" + msg + ", next=" + next + + ", errMsg=" + (err != null ? err.getMessage() : "N/A") + ']'); else if (log.isDebugEnabled()) log.debug("Failed to send message to next node [msg=" + msg + ", next=" + next + ", errMsg=" + (err != null ? err.getMessage() : "N/A") + ']'); @@ -4025,7 +4031,7 @@ class ServerImpl extends TcpDiscoveryImpl { /** * Check the last time a heartbeat message received. If the time is bigger than {@code hbCheckTimeout} than - * {@link TcpDiscoveryStatusCheckMessage} is sent accros the ring. + * {@link TcpDiscoveryStatusCheckMessage} is sent across the ring. */ private void checkHeartbeatsReceiving() { if (lastTimeStatusMsgSent < locNode.lastUpdateTime()) @@ -4045,11 +4051,9 @@ class ServerImpl extends TcpDiscoveryImpl { * Check connection aliveness status. */ private void checkConnection() { - if (!spi.failureDetectionTimeoutEnabled()) - return; - - if (!failureThresholdReached && U.currentTimeMillis() - locNode.lastDataReceivedTime() - >= spi.failureDetectionTimeout() && ring.hasRemoteNodes() && spiStateCopy() == CONNECTED) { + if (spi.failureDetectionTimeoutEnabled() && !failureThresholdReached && + U.currentTimeMillis() - locNode.lastDataReceivedTime() >= connCheckThreshold && + ring.hasRemoteNodes() && spiStateCopy() == CONNECTED) { log.info("Local node seems to be disconnected from topology (failure detection timeout " + "is reached): [failureDetectionTimeout=" + spi.failureDetectionTimeout() +
