You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Semyon Danilov (Jira)" <ji...@apache.org> on 2021/03/31 07:58:00 UTC
[jira] [Updated] (IGNITE-14448) Failure to connect to node leads to
hanging connection future if paired connections are used
[ https://issues.apache.org/jira/browse/IGNITE-14448?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Semyon Danilov updated IGNITE-14448:
------------------------------------
Description:
{code:java}
if ((CommunicationSpi<?>)spi instanceof TcpCommunicationSpi)
getTcpCommunicationSpi().setConnectionRequestor(invConnHandler);
if (connRequestor != null) {
...
if (isPairedConnection(node, tcpCommSpi))
throw new IgniteSpiException("Inverse connection protocol doesn't support paired connections");{code}
Turns out this exception is not handled property and connection future is never done. Then, striped pool threads wait forever on reserveClient() and cluster grinds to halt.This happens in versions which have communication-via-discovery and when usePairedConnections=true.
{code:java}
[12:06:18,110][SEVERE][sys-stripe-0-#1][TcpCommunicationSpi] Failed to send message to remote node [node=TcpDiscoveryNode [id=54ddcf8b-3e41-4efe-bb9d-8a0369e7b893, consistentId=54ddcf8b-3e4
1-4efe-bb9d-8a0369e7b893, addrs=ArrayList [127.0.0.1, 172.22.229.21], sockAddrs=HashSet [/127.0.0.1:0, ip-172-22-229-21.ec2.internal/172.22.229.21:0], discPort=0, order=47, intOrder=47, lastExchangeTime=1603983940522, loc=false, ver=8.7.25#20200910-sha1:b580d9fd, isClient=true], msg=GridIoMessage [plc=2, topic=TOPIC_CACHE, topicOrd=8, ordered=false, timeout=0, skipOnTimeout=f
alse, msg=GridDhtAtomicSingleUpdateRequest [key=KeyCacheObjectImpl [part=24, val=23576, hasValBytes=true], val=com.dream11.ignite.model.GetRoundSummaryRes [idHash=69226443, hash=580815760,roundId=23576, dataSource=MYSQL, sparkJobStatus=COMPLETED], prevVal=null, super=GridDhtAtomicAbstractUpdateRequest [onRes=false, nearNodeId=null, nearFutId=0, flags=near]], connIdx=-1]]
class org.apache.ignite.spi.IgniteSpiException: Inverse connection protocol doesn't support paired connections
at org.apache.ignite.internal.managers.communication.GridIoManager$TcpCommunicationInverseConnectionHandler.request(GridIoManager.java:3564)
at org.apache.ignite.spi.communication.tcp.internal.ConnectionClientPool.handleUnreachableNodeException(ConnectionClientPool.java:365)
at org.apache.ignite.spi.communication.tcp.internal.ConnectionClientPool.reserveClient(ConnectionClientPool.java:256)
at org.apache.ignite.spi.communication.tcp.TcpCommunicationSpi.sendMessage0(TcpCommunicationSpi.java:1132)
at org.apache.ignite.spi.communication.tcp.TcpCommunicationSpi.sendMessage(TcpCommunicationSpi.java:1083)
at org.apache.ignite.internal.managers.communication.GridIoManager.send(GridIoManager.java:1814)
at org.apache.ignite.internal.managers.communication.GridIoManager.sendToGridTopic(GridIoManager.java:1930)
at org.apache.ignite.internal.processors.cache.GridCacheIoManager.send(GridCacheIoManager.java:1257)
at org.apache.ignite.internal.processors.cache.GridCacheIoManager.send(GridCacheIoManager.java:1296)
at org.apache.ignite.internal.processors.cache.distributed.dht.atomic.GridDhtAtomicAbstractUpdateFuture.sendDhtRequests(GridDhtAtomicAbstractUpdateFuture
{code}
was:
{{if ((CommunicationSpi<?>)spi instanceof TcpCommunicationSpi)
getTcpCommunicationSpi().setConnectionRequestor(invConnHandler);
if (connRequestor != null) {
...
if (isPairedConnection(node, tcpCommSpi))
throw new IgniteSpiException("Inverse connection protocol doesn't support paired connections");}}
Turns out this exception is not handled property and connection future is never done. Then, striped pool threads wait forever on reserveClient() and cluster grinds to halt.
This happens in versions which have communication-via-discovery and when usePairedConnections=true.
{{[12:06:18,110][SEVERE][sys-stripe-0-#1][TcpCommunicationSpi] Failed to send message to remote node [node=TcpDiscoveryNode [id=54ddcf8b-3e41-4efe-bb9d-8a0369e7b893, consistentId=54ddcf8b-3e4
1-4efe-bb9d-8a0369e7b893, addrs=ArrayList [127.0.0.1, 172.22.229.21], sockAddrs=HashSet [/127.0.0.1:0, ip-172-22-229-21.ec2.internal/172.22.229.21:0], discPort=0, order=47, intOrder=47, lastExchangeTime=1603983940522, loc=false, ver=8.7.25#20200910-sha1:b580d9fd, isClient=true], msg=GridIoMessage [plc=2, topic=TOPIC_CACHE, topicOrd=8, ordered=false, timeout=0, skipOnTimeout=f
alse, msg=GridDhtAtomicSingleUpdateRequest [key=KeyCacheObjectImpl [part=24, val=23576, hasValBytes=true], val=com.dream11.ignite.model.GetRoundSummaryRes [idHash=69226443, hash=580815760,roundId=23576, dataSource=MYSQL, sparkJobStatus=COMPLETED], prevVal=null, super=GridDhtAtomicAbstractUpdateRequest [onRes=false, nearNodeId=null, nearFutId=0, flags=near]], connIdx=-1]]
class org.apache.ignite.spi.IgniteSpiException: Inverse connection protocol doesn't support paired connections
at org.apache.ignite.internal.managers.communication.GridIoManager$TcpCommunicationInverseConnectionHandler.request(GridIoManager.java:3564)
at org.apache.ignite.spi.communication.tcp.internal.ConnectionClientPool.handleUnreachableNodeException(ConnectionClientPool.java:365)
at org.apache.ignite.spi.communication.tcp.internal.ConnectionClientPool.reserveClient(ConnectionClientPool.java:256)
at org.apache.ignite.spi.communication.tcp.TcpCommunicationSpi.sendMessage0(TcpCommunicationSpi.java:1132)
at org.apache.ignite.spi.communication.tcp.TcpCommunicationSpi.sendMessage(TcpCommunicationSpi.java:1083)
at org.apache.ignite.internal.managers.communication.GridIoManager.send(GridIoManager.java:1814)
at org.apache.ignite.internal.managers.communication.GridIoManager.sendToGridTopic(GridIoManager.java:1930)
at org.apache.ignite.internal.processors.cache.GridCacheIoManager.send(GridCacheIoManager.java:1257)
at org.apache.ignite.internal.processors.cache.GridCacheIoManager.send(GridCacheIoManager.java:1296)
at org.apache.ignite.internal.processors.cache.distributed.dht.atomic.GridDhtAtomicAbstractUpdateFuture.sendDhtRequests(GridDhtAtomicAbstractUpdateFuture}}
> Failure to connect to node leads to hanging connection future if paired connections are used
> --------------------------------------------------------------------------------------------
>
> Key: IGNITE-14448
> URL: https://issues.apache.org/jira/browse/IGNITE-14448
> Project: Ignite
> Issue Type: Bug
> Components: networking
> Affects Versions: 2.10
> Reporter: Semyon Danilov
> Assignee: Semyon Danilov
> Priority: Major
>
> {code:java}
> if ((CommunicationSpi<?>)spi instanceof TcpCommunicationSpi)
> getTcpCommunicationSpi().setConnectionRequestor(invConnHandler);
> if (connRequestor != null) {
> ...
> if (isPairedConnection(node, tcpCommSpi))
> throw new IgniteSpiException("Inverse connection protocol doesn't support paired connections");{code}
> Turns out this exception is not handled property and connection future is never done. Then, striped pool threads wait forever on reserveClient() and cluster grinds to halt.This happens in versions which have communication-via-discovery and when usePairedConnections=true.
> {code:java}
> [12:06:18,110][SEVERE][sys-stripe-0-#1][TcpCommunicationSpi] Failed to send message to remote node [node=TcpDiscoveryNode [id=54ddcf8b-3e41-4efe-bb9d-8a0369e7b893, consistentId=54ddcf8b-3e4
> 1-4efe-bb9d-8a0369e7b893, addrs=ArrayList [127.0.0.1, 172.22.229.21], sockAddrs=HashSet [/127.0.0.1:0, ip-172-22-229-21.ec2.internal/172.22.229.21:0], discPort=0, order=47, intOrder=47, lastExchangeTime=1603983940522, loc=false, ver=8.7.25#20200910-sha1:b580d9fd, isClient=true], msg=GridIoMessage [plc=2, topic=TOPIC_CACHE, topicOrd=8, ordered=false, timeout=0, skipOnTimeout=f
> alse, msg=GridDhtAtomicSingleUpdateRequest [key=KeyCacheObjectImpl [part=24, val=23576, hasValBytes=true], val=com.dream11.ignite.model.GetRoundSummaryRes [idHash=69226443, hash=580815760,roundId=23576, dataSource=MYSQL, sparkJobStatus=COMPLETED], prevVal=null, super=GridDhtAtomicAbstractUpdateRequest [onRes=false, nearNodeId=null, nearFutId=0, flags=near]], connIdx=-1]]
> class org.apache.ignite.spi.IgniteSpiException: Inverse connection protocol doesn't support paired connections
> at org.apache.ignite.internal.managers.communication.GridIoManager$TcpCommunicationInverseConnectionHandler.request(GridIoManager.java:3564)
> at org.apache.ignite.spi.communication.tcp.internal.ConnectionClientPool.handleUnreachableNodeException(ConnectionClientPool.java:365)
> at org.apache.ignite.spi.communication.tcp.internal.ConnectionClientPool.reserveClient(ConnectionClientPool.java:256)
> at org.apache.ignite.spi.communication.tcp.TcpCommunicationSpi.sendMessage0(TcpCommunicationSpi.java:1132)
> at org.apache.ignite.spi.communication.tcp.TcpCommunicationSpi.sendMessage(TcpCommunicationSpi.java:1083)
> at org.apache.ignite.internal.managers.communication.GridIoManager.send(GridIoManager.java:1814)
> at org.apache.ignite.internal.managers.communication.GridIoManager.sendToGridTopic(GridIoManager.java:1930)
> at org.apache.ignite.internal.processors.cache.GridCacheIoManager.send(GridCacheIoManager.java:1257)
> at org.apache.ignite.internal.processors.cache.GridCacheIoManager.send(GridCacheIoManager.java:1296)
> at org.apache.ignite.internal.processors.cache.distributed.dht.atomic.GridDhtAtomicAbstractUpdateFuture.sendDhtRequests(GridDhtAtomicAbstractUpdateFuture
> {code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)