[
https://issues.apache.org/jira/browse/IGNITE-15767?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ivan Bessonov updated IGNITE-15767:
-----------------------------------
Description:
See [https://bugs.openjdk.java.net/browse/JDK-8247750]
ServerSocket.accept() with no timeout may throw SocketTimeoutException when the
process receives a signal. It can cause unexpected exception in the disco
reader:
{noformat}
[09:52:26,301][SEVERE][tcp-disco-srvr-[:47500]-#3-#71][TcpDiscoverySpi] Failed
to accept TCP
connection.[09:52:26,301][SEVERE][tcp-disco-srvr-[:47500]-#3-#71][TcpDiscoverySpi]
Failed to accept TCP connection.java.net.SocketTimeoutException: Accept timed
out at java.base/java.net.PlainSocketImpl.socketAccept(Native Method)
at
java.base/java.net.AbstractPlainSocketImpl.accept(AbstractPlainSocketImpl.java:458)
at java.base/java.net.ServerSocket.implAccept(ServerSocket.java:565)
at java.base/java.net.ServerSocket.accept(ServerSocket.java:533)
at
org.apache.ignite.spi.discovery.tcp.ServerImpl$TcpServer.body(ServerImpl.java:6750)
at org.apache.ignite.internal.util.worker.GridWorker.run(GridWorker.java:119)
at
org.apache.ignite.spi.discovery.tcp.ServerImpl$TcpServerThread.body(ServerImpl.java:6673)
at org.apache.ignite.spi.IgniteSpiThread.run(IgniteSpiThread.java:57){noformat}
There are mentions of this on StackOverflow and userlist, and we also see it
when running Docker on Alpine 3.15. Speculation is that this is caused by a
combination of a specific version of Linux kernel and environment.
Sidenote: based on strace analysis on Alpine, it doesn't even receive any
signals; it could be so that Alpine interrupts the syscall "as if" by a signal.
The bug is not fixed in JDK 11 (surprisingly). WA is easy though - wrap the
accept in try-catch, and retry if getting the unexpected timeout exception.
was:
See [https://bugs.openjdk.java.net/browse/JDK-8247750]
ServerSocket.accept() with no timeout may throw SocketTimeoutException when the
process receives a signal. It can cause unexpected exception in the disco
reader:
{noformat}
[09:52:26,301][SEVERE][tcp-disco-srvr-[:47500]-#3-#71][TcpDiscoverySpi] Failed
to accept TCP
connection.[09:52:26,301][SEVERE][tcp-disco-srvr-[:47500]-#3-#71][TcpDiscoverySpi]
Failed to accept TCP connection.java.net.SocketTimeoutException: Accept timed
out at java.base/java.net.PlainSocketImpl.socketAccept(Native Method) at
java.base/java.net.AbstractPlainSocketImpl.accept(AbstractPlainSocketImpl.java:458)
at java.base/java.net.ServerSocket.implAccept(ServerSocket.java:565) at
java.base/java.net.ServerSocket.accept(ServerSocket.java:533) at
org.apache.ignite.spi.discovery.tcp.ServerImpl$TcpServer.body(ServerImpl.java:6750)
at org.apache.ignite.internal.util.worker.GridWorker.run(GridWorker.java:119)
at
org.apache.ignite.spi.discovery.tcp.ServerImpl$TcpServerThread.body(ServerImpl.java:6673)
at org.apache.ignite.spi.IgniteSpiThread.run(IgniteSpiThread.java:57){noformat}
{{}}
There are mentions of this on StackOverflow and userlist, and we also see it
when running Docker on Alpine 3.15. Speculation is that this is caused by a
combination of a specific version of Linux kernel and environment.
Sidenote: based on strace analysis on Alpine, it doesn't even receive any
signals; it could be so that Alpine interrupts the syscall "as if" by a signal.
The bug is not fixed in JDK 11 (surprisingly). WA is easy though - wrap the
accept in try-catch, and retry if getting the unexpected timeout exception.
> Need to workaround JDK bug JDK-8247750
> --------------------------------------
>
> Key: IGNITE-15767
> URL: https://issues.apache.org/jira/browse/IGNITE-15767
> Project: Ignite
> Issue Type: Improvement
> Reporter: Ivan Bessonov
> Priority: Major
> Fix For: 2.13
>
>
> See [https://bugs.openjdk.java.net/browse/JDK-8247750]
> ServerSocket.accept() with no timeout may throw SocketTimeoutException when
> the process receives a signal. It can cause unexpected exception in the disco
> reader:
> {noformat}
> [09:52:26,301][SEVERE][tcp-disco-srvr-[:47500]-#3-#71][TcpDiscoverySpi]
> Failed to accept TCP
> connection.[09:52:26,301][SEVERE][tcp-disco-srvr-[:47500]-#3-#71][TcpDiscoverySpi]
> Failed to accept TCP connection.java.net.SocketTimeoutException: Accept
> timed out at java.base/java.net.PlainSocketImpl.socketAccept(Native Method)
> at
> java.base/java.net.AbstractPlainSocketImpl.accept(AbstractPlainSocketImpl.java:458)
> at java.base/java.net.ServerSocket.implAccept(ServerSocket.java:565)
> at java.base/java.net.ServerSocket.accept(ServerSocket.java:533)
> at
> org.apache.ignite.spi.discovery.tcp.ServerImpl$TcpServer.body(ServerImpl.java:6750)
> at org.apache.ignite.internal.util.worker.GridWorker.run(GridWorker.java:119)
> at
> org.apache.ignite.spi.discovery.tcp.ServerImpl$TcpServerThread.body(ServerImpl.java:6673)
> at
> org.apache.ignite.spi.IgniteSpiThread.run(IgniteSpiThread.java:57){noformat}
> There are mentions of this on StackOverflow and userlist, and we also see it
> when running Docker on Alpine 3.15. Speculation is that this is caused by a
> combination of a specific version of Linux kernel and environment.
> Sidenote: based on strace analysis on Alpine, it doesn't even receive any
> signals; it could be so that Alpine interrupts the syscall "as if" by a
> signal.
> The bug is not fixed in JDK 11 (surprisingly). WA is easy though - wrap the
> accept in try-catch, and retry if getting the unexpected timeout exception.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
