Exception: Network unreachable

[dzus]

I have started a TS 2 server on my router which is a part of big wireless community network in Prague. And sometimes I get this message in the server.log file:

21-01-04 04:02:06,ERROR,All,TTSUDPSender.Execute, SID: 1 Exception1 EIdSocketError.Socket Error # 101
Network is unreachable.
21-01-04 04:02:07,ERROR,All,TTSUDPSender.Execute, SID: 1 Exception2 EIdSocketError.Socket Error # 101
Network is unreachable.

And then it's not posible to connect to the TS2 server.

Important note:
because of the size of the network we are using dynamic routing protocols like OSPF and BGP and on the router is running the Zebra routing daemon.
If any of the links fails, then the corresponding route disappears from the routing table. When the routing daemon is killed, then all routes (except routes to local interfaces) including default route are deleted from RT.
This is a normal behavior and I see no reason why this should corrupt functionality of any userspace program (including TS2).

Server version is 2.0.19.40

I hope, that you will be able to fix this in the (near :) ) future.

[Dummer Sack]

Hmm, 101 Is normaly not the error code for Network unreachable. Strange.

Anyway I think the reason is that the default route is deleted. Since the server and client are sending continously a heartbeat signal they will get Network Unreachable for the time the default route is down.


Here a short description of the Network is unreachable error:
WSAENETUNREACH (10051) Network is unreachable.

Berkeley description: A socket operation was attempted to an unreachable network.

WinSock description: Almost same as Berkeley. For WinSock, this error is equivalent to Berkeley's EHOSTUNREACH error, the catch-all error for unreachable hosts. "You can't get there from here."

TCP/IP scenario: The local network system could generate this error if there isn't a default route configured. Typically, though, WinSock generates this error when it receives a "host unreachable" ICMP message from a router. The ICMP message means that a router can't forward the IP datagram, possibly because it didn't get a response to the ARP request (which might mean the destination host is down). Note: this error may also result if you are trying to send a multicast packet and the default gateway does not support multicast (check your interface configuration).

User suggestions: Try to ping the destination host, to see if you get the same results (chances are, you will). Check the destination address itself; is it the one you wanted to go to? Check whether you have a router configured in your network system (your WinSock implementation). Do a traceroute to try to determine where the failure occurs along the route between your host and the destination host.

WinSock functions: connect(), sendto(), FD_CONNECT

Additional functions: Any function that does network I/O: recv(), recvfrom(), send(), FD_READ, FD_WRITE

See also: WSAEHOSTUNREACH

[dzus]

Originally posted by Dummer Sack
Anyway I think the reason is that the default route is deleted. Since the server and client are sending continously a heartbeat signal they will get Network Unreachable for the time the default route is down.


As I said any of the routes can be deleted if there is no router propagating this route through OSPF (including default route). It has good reason (it is used to switch routing to a backup links).

But I have no idea why broken connection with a client should stop the whole TS server - after this error I must manually restart the server!
And anyway I think that at 4 am, there were no clients connected to the server.

[Dummer Sack]

Ah now I know what you mean.
You mean that the TS server will not reestabish the connection after Network Unreachable and is not accepting any new ones.
Yes, that seems to be a bug.

Can you try if that still happens with the b50 server (you can get it in the developer releases forum)?
I want to be sure that the bug is still there before I file a bug report.

[dzus]

Originally posted by Dummer Sack
Ah now I know what you mean.
You mean that the TS server will not reestabish the connection after Network Unreachable and is not accepting any new ones.


Yes, exactly - I'm sorry that I forgotten to write it in my first message.

Originally posted by Dummer Sack

Can you try if that still happens with the b50 server (you can get it in the developer releases forum)?


I'll try it ...

[dzus]

I tried the b50 version of the server and it crashes in the same way as 2.0.19.40.
You can do a simple test - try to delete default route on any server running TS and see what happens :)

I think that you can send a bug report because it is definitely a bug.

[Dummer Sack]

Ok, I have put it in mantis.

http://mantis.teamspeak.org/bug_view_page.php?bug_id=0000586