The SFTP Connection are not closed when an exception is thrown using 3GPP XML Collector
Description
As we are using the direct sftp.3gpp we have seen multiple tcp ports opened without closing. This is majorly happens when no XML files are available for collection.
For e.g. we are able to see 900 ssh connections similar to something like below:
tcp 0 0 rchsrv-opennms01:56522 ::ffff:192.18.239.156:ssh ESTABLISHED 28341/java
All of them went away at OpenNMS side, when we restarted OpenNMS process.
May be, OpenNMS side TCP layer is not closing the sockets (If xml files are not available for collection in the destination node, after a particular waiting period we may need to close the opened sftp connection sockets).
So, the open files issue at OpenNMS side may be probably due to so many sftp connections that it keeps open.
Is there any workaround for ensuring sftp sockets in OpenNMS don’t hang with long timeouts.
Acceptance / Success Criteria
None
Lucidchart Diagrams
Activity
Show:
Alejandro Galue May 22, 2014 at 2:33 PM
Nikhil, I've merged the fix on revision fab1a8a0e9c42574a850e3340e935d465949cac9 for space/14.1
Nikhil Tambe May 22, 2014 at 1:06 PM
Can this be merged in Space branch as well
Alejandro Galue May 20, 2014 at 10:37 AM
Merged into master for 1.13.2 on revision 03cb49047a8dc2ed0d04a1ba8f142ce3e897a609.
The VTD-XML implementation in 1.13 has been updated as well.
Alejandro Galue May 20, 2014 at 10:24 AM
Fixed on revision d128fd4a09fd47f92c6a14f257b0fe6c517994c8 for 1.12.
It is guaranteed now that all the opened connections (and input streams) are going to be closed even if an exception is thrown while processing any kind of data (not only 3GPP through SFTP).
As we are using the direct sftp.3gpp we have seen multiple tcp ports opened without closing. This is majorly happens when no XML files are available for collection.
For e.g. we are able to see 900 ssh connections similar to something like below:
tcp 0 0 rchsrv-opennms01:56522 ::ffff:192.18.239.156:ssh ESTABLISHED 28341/java
All of them went away at OpenNMS side, when we restarted OpenNMS process.
May be, OpenNMS side TCP layer is not closing the sockets (If xml files are not available for collection in the destination node, after a particular waiting period we may need to close the opened sftp connection sockets).
So, the open files issue at OpenNMS side may be probably due to so many sftp connections that it keeps open.
Is there any workaround for ensuring sftp sockets in OpenNMS don’t hang with long timeouts.