Incorrect Client Commandline Picked up for Packetbeat flows #9151
Assignees
Comments
adriansr
added a commit
to adriansr/beats
that referenced
this issue
Dec 7, 2018
The current implementation of Packetbeat's process monitor detects the local process by looking at the local port number used. However, differnet processes can be bound to the same port on different interfaces. This patch fixes the problem by looking up the process by using the tuple (address, port). Fixes elastic#9151
adriansr
added a commit
that referenced
this issue
Dec 11, 2018
The current implementation of Packetbeat's process monitor detects the local process by looking at the local port number used. However, different processes can be bound to the same port on different interfaces. This patch fixes the problem by looking up the process by using the tuple (address, port). There's still a case where Packetbeat can pick the wrong process: When one process is bound to INADDR_ANY and another to a specific local address. Testing suggests that the last socket to be bound takes precedence over the other. However, I couldn't find this behavior documented anywhere. As Packetbeat can't tell which socket was bound first, there's no way of telling which process is really receiving the traffic. This PR will give precedence to a socket bound to a local IP-address over a socket bound to INADDR_ANY. Fixes #9151
adriansr
added a commit
to adriansr/beats
that referenced
this issue
Dec 11, 2018
The current implementation of Packetbeat's process monitor detects the local process by looking at the local port number used. However, different processes can be bound to the same port on different interfaces. This patch fixes the problem by looking up the process by using the tuple (address, port). There's still a case where Packetbeat can pick the wrong process: When one process is bound to INADDR_ANY and another to a specific local address. Testing suggests that the last socket to be bound takes precedence over the other. However, I couldn't find this behavior documented anywhere. As Packetbeat can't tell which socket was bound first, there's no way of telling which process is really receiving the traffic. This PR will give precedence to a socket bound to a local IP-address over a socket bound to INADDR_ANY. Fixes elastic#9151 (cherry picked from commit 824d443)
adriansr
added a commit
that referenced
this issue
Dec 12, 2018
The current implementation of Packetbeat's process monitor detects the local process by looking at the local port number used. However, different processes can be bound to the same port on different interfaces. This patch fixes the problem by looking up the process by using the tuple (address, port). There's still a case where Packetbeat can pick the wrong process: When one process is bound to INADDR_ANY and another to a specific local address. Testing suggests that the last socket to be bound takes precedence over the other. However, I couldn't find this behavior documented anywhere. As Packetbeat can't tell which socket was bound first, there's no way of telling which process is really receiving the traffic. This PR will give precedence to a socket bound to a local IP-address over a socket bound to INADDR_ANY. Fixes #9151 (cherry picked from commit 824d443)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Packetbeat is occasionally picking up a random (incorrect) client_commandline process in flow data. The process has nothing to do with the related flow.
The text was updated successfully, but these errors were encountered: