Cannot use container's name as --server-addr in configuration.
#773
Comments
|
So, what does |
|
|
This is the config file: {
"server": "con_kcp",
"server_port": 1090,
"local_address": "0.0.0.0",
"local_port": 1080,
"password": "mypass",
"timeout":120,
"method":"chacha20-ietf-poly1305",
"fast_open": false,
"log": {
"level": 1
}
}
|
|
What did you see from logs? Try to set |
|
Name resolution seems to take no effect: |
|
You have to make a request. The remote server will only be resolved when creating new tunnels. |
|
Well, my log dumps as: The ping con_kcp
PING con_kcp (172.20.0.3): 56 data bytes
64 bytes from 172.20.0.3: seq=0 ttl=64 time=0.322 ms
64 bytes from 172.20.0.3: seq=1 ttl=64 time=0.431 msEverything works fine after changing the server address from using the domain name to IP address: |
Yes. It didn't show the actual resolved IP in log. :( Is the This line of How to reproduce? Could I reproduce it locally with my laptop? |
|
I have narrowed the scope of this issue to the problem of whether a domain name is valid or not. For a simple POC, I use the docker compose file and configuration files of ss: ssserver config file {
"server": "0.0.0.0",
"server_port": 1090,
"password": "mypass",
"timeout": 120,
"method": "chacha20-ietf-poly1305",
"fast_open": false,
"log": {
"level": 3
}
}sslocal config file {
"server": "server.ss.invalid",
"server_port": 1090,
"local_address": "0.0.0.0",
"local_port": 1080,
"password": "mypass",
"timeout":120,
"method":"chacha20-ietf-poly1305",
"fast_open": false,
"log": {
"level": 3
}
}docker compose file: version: '3.5'
services:
server.ss.invalid:
image: ghcr.io/shadowsocks/ssserver-rust:latest
container_name: server.ss.invalid
restart: always
ports:
- 127.0.0.1:1090:1090
volumes:
- type: bind
source: /tmp/poc/server.json
target: /etc/shadowsocks-rust/config.json
read_only: true
local.ss.invalid:
image: ghcr.io/shadowsocks/sslocal-rust:latest
container_name: local.ss.invalid
restart: always
ports:
- 127.0.0.1:1080:1080
volumes:
- type: bind
source: /tmp/poc/client.json
target: /etc/shadowsocks-rust/config.json
read_only: true
I think this problem is related to the verification of domain name format which is done before the actual resolution. For container circumstances, the container name, which is a superset of valid domain names, acts as a domain name in most situations. |
|
It's irrelevent. This change is only for |
|
Insterestingly, Run So there was a DNS query sent to There is no packet could be captured by |
|
DNS server address for containers which use custom network is 127.0.0.11:53. Container Try to capture packets on lo interface to see if any clue can be found. |
Tried and captured nothing. |
|
:( sad... |
Sending queries directly to the But it is actually working well for other valid domains, like: So I think the When sending queries to
|
|
|
|
Everything works fine after changing containers' name to version: '3.5'
services:
server.ss:
image: ghcr.io/shadowsocks/ssserver-rust:latest
container_name: server.ss
restart: always
ports:
- 127.0.0.1:1090:1090
volumes:
- type: bind
source: /tmp/poc/server.json
target: /etc/shadowsocks-rust/config.json
read_only: true
local.ss:
image: ghcr.io/shadowsocks/sslocal-rust:latest
container_name: local.ss
restart: always
ports:
- 127.0.0.1:1080:1080
volumes:
- type: bind
source: /tmp/poc/client.json
target: /etc/shadowsocks-rust/config.json
read_only: trueBut why Well, |
Yes, I have mentioned that earlier. That is why I used Should we opt to use some more 'naive' name resolution method before the action that is taken by |
|
It can be done by removing |
|
Maybe, providing a fallback strategy is more compatible. For domain names that are obviously not valid, or domain names that cannot be handled by |
|
https://github.com/shadowsocks/shadowsocks-rust/releases/tag/v1.13.4 Add a |
|
Seems still not working. I have updated my client config in which "DNS" is now pointed to "system": {
"server": "server.ss.invalid",
"server_port": 1090,
"local_address": "0.0.0.0",
"local_port": 1080,
"password": "mypass",
"timeout":120,
"method":"chacha20-ietf-poly1305",
"fast_open": false,
"dns": "system",
"log": {
"level": 3
}
}NOTE: Don't test by accessing "baidu.com" in non-incognito mode. The web seems to be cached and shows the result correctly. |
This comment was marked as outdated.
This comment was marked as outdated.
|
Set |
打扰一下，我觉得这样做，并不太优雅。为什么不能先做hosts文件解析，没结果再扔应用定义的dns服务做处理？ |
|
It doesn't related to host file. |
|
It works now. Close this issue. |
|
BTW, why not just make the container name to be a valid domain name? |
|
Well, the main reason is that we haven't encountered such a problem before by using Shifting from old container names to new ones requires some time, and we are now making the process to do that. |
I use
kcptunandsslocaltogether to speed up network access. Both of these two services run in containers (two separate containers).I can ping
kcpserver fromssby the container name, saycon_kcp.However, using this
con_kcpas--server-addrin the configure file fails the connection to kcp.It works fine to use container's address instead of
con_kcpdomain name.How could I resolve this?
The text was updated successfully, but these errors were encountered: