gsliepen / tinc

a VPN daemon
http://tinc-vpn.org/
Other
1.87k stars 280 forks source link

Really need help #419

Open ptorrent opened 1 year ago

ptorrent commented 1 year ago

Hello,

I've a tinc network ~1000 nodes

we can test on 1.0.36, on 1.1pre18

It's not working as expected, there is a lot of issue like

1) 1.1pre18 ConnectTo 1.1pre18 with TunnelServer (i'm able to ping but I never receive the subnet up event server side) (not able to ping any others nodes on this server even if I've host file and other node are connected on the same server as me.

2) Infinity loop subnet up / down without TunnelServer on 1.1pre18 / 1.1pre16

3) What's the real use case of using "TunnelServer". If I get it correctly you can only access to the host set in ConnectTo?

4) I would love to have a p2p VPN, and I've a point-to-point VPN that doesn't work properly .

5) If I don't set "TunnelServer" I receive subnet-up events from node with tinc turned off... How it can be possible ? on 1.1pre18/ 1.1pre16

6) What the difference between StrictSubnet and TunnelServer ?

7) If I set StrictSubnet = no and TunnelServer = yes , looks better but does it make sense ? Forget about it... I've subnet up but unable to ping any hosts other than ConnectTo...

8) Is there something wrong with UDP ? Even with UDPDiscovery = no and TCPOnly I receive UDP paquet

image

9) how can I have subnet up from node offline ? Some node are connecting in loop event if they are off (no tinc running on the device). But if I do

tinc --pidfile="/etc/tinc/xxxx/process.pid" disconnect NODE_OFFLINE_BUT_SUBNET_UP

After that it's ok. Should I do disconnect NODE after a subnet down ?

10) segfault after a tinc purge: (1.1pre18)

tincd[18701]: segfault at 55c324000000 ip 00007f4914663544 sp 00007ffd6c5fae38 error 4 in libc-2.24.so[7f49145e8000+195000]

What is doing a purge ? With debug level 3, no more log sorry

11) After doing a puge, I've this log (regarding debug level3 output)

Got DEL_SUBNET from NODE_XXXXXX ([public ip] port 1306) for NODE_YYYYYYY which is not in our node tree

What does this mean? Do I've to call a purge some times ?

12) Does it make sense to put "Weight = 0" for a node acting as server ?

I would really appreciate your help, you can contact me. We can pay your time. I can't help the project otherwise and I think this will help a lot the project...

ptorrent commented 1 year ago

Other point.

This is my scenario

NODE_A + NODE_B + NODE_C ===> ConnectTo = NODE_X

NODE_A is running on 1.1pre18 NODE_B + C are running on 1.0.36

There is no address set in NODE_A, NODE_B and NODE_C host file.

image

By running tinc dump edges:

NODE_A to NODE_B at [PUBLIE_IP_NODE_B]  port 655 local [PUBLIE_IP_NODE_A]  port 655 options b weight 82
NODE_A  to NODE_C at [PUBLIE_IP_NODE_C]  port 655 local [PUBLIE_IP_NODE_A]  port 655 options b weight 74
NODE_B to NODE_A at [PUBLIE_IP_NODE_A]  port 655 local 192.168.1.114 port 655 options b weight 82
NODE_C  to NODE_A at [PUBLIE_IP_NODE_A] port 655 local 192.168.1.130 port 655 options b weight 74

How it's possible that NODE_B + NODE_C are connected to NODE_A ? Is it because NODE_X and NODE_A are sharing the same modem ? Something with UDP ?

I've the feeling that when NODE_A is disconnected all nodes connected to him are disconnected too... But NODE_A is not acting like a server (no up 100% of the time)

ptorrent commented 1 year ago

Other point.

This is my scenario

NODE_A + NODE_B + NODE_C ===> ConnectTo = NODE_X

NODE_A is running on 1.1pre18 NODE_B + C are running on 1.0.36

There is no address set in NODE_A, NODE_B and NODE_C host file.

image

By running tinc dump edges:

NODE_A to NODE_B at [PUBLIE_IP_NODE_B]  port 655 local [PUBLIE_IP_NODE_A]  port 655 options b weight 82
NODE_A  to NODE_C at [PUBLIE_IP_NODE_C]  port 655 local [PUBLIE_IP_NODE_A]  port 655 options b weight 74
NODE_B to NODE_A at [PUBLIE_IP_NODE_A]  port 655 local 192.168.1.114 port 655 options b weight 82
NODE_C  to NODE_A at [PUBLIE_IP_NODE_A] port 655 local 192.168.1.130 port 655 options b weight 74

How it's possible that NODE_B + NODE_C are connected to NODE_A ? Is it because NODE_X and NODE_A are sharing the same modem ? Something with UDP ?

I've the feeling that when NODE_A is disconnected all nodes connected to him are disconnected too... But NODE_A is not acting like a server (no up 100% of the time)

solved by blocking 655 port tcp/udp on the NODE_A....