[
https://issues.apache.org/jira/browse/SSHD-908?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16810035#comment-16810035
]
Goldstein Lyor commented on SSHD-908:
-------------------------------------
The leak seems obvious to me
- the {{SshClient}} is created-started-stopped at a very high rate. While we
have invested a lot of effort into trying to allow SshClient/Server(s) to be
re-created/started I strongly advise against it. A Java application should use
*one* instance (client or server or one of each) and use it to create as many
sessions as necessary. Therefore the use-case depicted in code seems a bit
unnatural.
- the {{sshSession}} is not properly disposed of
- ditto for the port forwarding
I understand this might be just some sample code, but if it is supposed to
detect leaks, at least it should be written in a manner as similar as possible
as real code - otherwise we will not be able to distinguish real leaks from
"synthetic" ones due to improper disposal of resources.
Here is how I would expect a program to look like (example is for client but
similar applies to server):
{code:java|title=Main application code sample}
SshClient client = ...setup and configure...
client.start();
while (...program is active...) {
try (ClientSession session = ...connect and auhenticate using the client
instance) {
....
session.startPortForwarding(....);
try {
....use the forwarding until no longer necessary...
....see also PortForwardingTracker that can be used as a
try-with-resource instead of this code block...
} finally {
session.stopPortForwarding();
}
}
}
client.stop();
{code}
> Leak of process handles caused by local port forwarding connections
> -------------------------------------------------------------------
>
> Key: SSHD-908
> URL: https://issues.apache.org/jira/browse/SSHD-908
> Project: MINA SSHD
> Issue Type: Bug
> Affects Versions: 2.1.0
> Reporter: Roberto Deandrea
> Priority: Major
> Attachments: repro.zip
>
>
> Hi,
> We have build an Apache SSHD Client (based on Apache SSHD 2.1.0) to forward
> local sockets to target servers.
> Our client is configured to use NIO2 builtin service provider, NIO2 Mina
> service provider and NIO2 Netty service provider.
> We have detected a leak of handles monitoring the java process with ALL above
> NIO2 service provider.
> When the limit of handles per process is reached, the java process become
> unresponsive.
>
> We suspect that the problem could be into Apache SSHD Client code.
> The problem is reproducible in our lab environment.
>
> Can you help us troubleshoot this problem and in that case what kind of info
> you need ?
> Did you see a similar problem before and if so, can you give us a solution ?
>
> Kind Regards
> Roberto Deandrea
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
