call home NOT initiating the connection after temporarily disconnection

[vishalwc]

Hi Michal

from Netconf server and TCP call home client shows the connection is established
NODEC21 ~]# netstat -anp | grep 4334
tcp 0 0 51.51.51.1:43364 192.168.2.15:4334 ESTABLISHED 3870/netopeer2-serv

from the netconf client and TCP server ( 192.168.2.15) there is no connection
[mgmt-380 ~]# netstat -anp | grep 4334
tcp6 0 0 :::4334 :::* LISTEN 896/java

please let me know what can be issue

[michalvasko]

Hi,
I will need more information than this to actually help you. What is exactly your scenario? I mean what configuration does your netopeer2 have (ietf-netconf-server data)? I guess you connected using Call Home and then how did the session terminate? Please, simply provide much more details.

Regards,
Michal

[vishalwc]

Hi Michal

Thanks for the feedback

The call home connection is persistence type
The connection was left for whole night and in between due to network condition it got temporarily disconnected .
later the call home client, even after the connection restore ,got stuck in the ESTABLISHED state and not initiating the call home request

TCP dump on 4334 port for that interface is not showing any packets or communication

[michalvasko]

Hi,
well, was the NETCONF session terminated? You would know by the output of the server and/or client, so can you please paste it here?

Regards,
Michal

[vishalwc]

yes the call home server cleaned connection from his end but that was after idetifying that connection is temporarily lost
so ssh connection was terminated form the server

[michalvasko]

Okay, so can you please paste it here? I need to know exactly what messages were printed so I know what code was executed.

Regards,
Michal

[vishalwc]

actually there are no logs that server disconnected or related to it , may be because the connection was lost temporarily it did not reach .
however after some time connected call home client should realize that connection is lost and retry the call home connection which is not happening.

[michalvasko]

Well, now I am confused. I first asked you whether the NETCONF session was terminated and you said that server did clean the connection but now you are saying there are no logs of that so how do you know? Anyway, I am just trying to find out if there is some other problem in libnetconf2 other than missing keep-alive feature. But, we enable keep-alive on TCP level so that should be enough, in theory.

Still, with these problems I need to know exactly what was happening on TCP layer (SSH) and on NETCONF layer, otherwise I am unable to help.

Regards,
Michal

[vishalwc]

Hi Michal

The call home server is listening on 4334 port and it cleaned the connection from the logs
but netopeer2 running on the call home client does not change the status and does not have any logs in that period .

Note : The mgmt server which is call home server in our case is cleaning the connection and earlier netstat show it .

what is the work around at TCP level as the connection in deployment will not come up in customer premises

[michalvasko]

Hi,
meaning NETCONF client disconnects but NETCONF server does not. Like I said, not having implemented Call Home keep-alive (on NETCONF layer) we rely on the TCP one. That does not seem to do its job so I will investigate more. However, I am working on other things currently so I will reply later.

Regards,
Michal

[vishalwc]

Thanks Michal appreciate your support .
will be helpful if I get feedback on work around by tomorrow

[michalvasko]

Hi,
then I must disappoint you because we have state holidays tomorrow and the day after, so I should be able to take a look next week, perhaps Monday.

Regards,
Michal

[vishalwc]

Hi Michal ,
did you get a chance to work on it .
I am planning to enable socket with keep alive message.

[michalvasko]

Hi,
I did just now and it is possible the keep-alive is working because, by default, an idle connection is dropped after ~2 hours and 11 minutes. However, it was set on client-side, which is probably why the server did not detect it. I have moved it to server-side and adjusted the options so that disconnected socket should be detected after ~1 minute. Can you test it with the latest libnetconf2, please? I was seeing the packets using tcpdump but not tried whether it really works and Call Home sessions are re-established.

Regards,
Michal

[vishalwc]

Thanks Michal

looks like you change to server however
there is common place nc_sock_connect() where the socket is created and connect call is made
we can have all session created with keep alive .

if(setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE, &optval, optlen) < 0) {
close(sock);
sock = -1;
continue;
}

while creating the socket isn't the right place ? please let me if I am missing something

Incidentally as the version which I am using is already being part of product so NOT sure I can take latest libnetconf2 library and keep the old netopeer2 same and it will work together .

appreciate your feedback earliest