Shadowsocks-libev: when using domain as server address, ss-libev would directly exit if the domain cannot be resolved

[zhou13]

For example, if I wrote

{
    "server":"example.com",
    ......
}

and if my network is not available (e.g., when booting), ss-server will directly exit with 255 and output

 2015-03-05 11:25:19 ERROR: getaddrinfo: Temporary failure in name resolution
 2015-03-05 11:25:19 ERROR: failed to resolve the provided hostname

I wish it shadowsocks would retry to resolve the domain until the network becomes available.

[zhou13]

BTW, the python version of shadowsocks works under this situation.

[madeye]

1. Use IP address instead.
2. Test the network before start shadowsocks.

[zhou13]

Is there any good reason not to retry when the first domain resolution fails?

The reason I do not use IP address is that my server contain both IPv4 and IPv6 address. I want to make shadowsocks use IPv6 if ipv6 network is available and otherwise use ipv4. Using IP address will lose this function.

The problem is that I want to start shadowsocks as a system daemon (in systemd), so it is not convenient for me to test the network before start shadowsocks. Although I can make systemd restart for me, it is not an elegant solution for me.

[perqin]

I recently encountered this issue on Arch Linux. The shadowsocks-libev systemd unit is started very first and network-online.target hasn't been pulled in. Using domain instead of IP address results in the error below:

Mar 17 18:01:31 perqin-nuc ss-local[485]:  2019-03-17 18:01:31 ERROR: getaddrinfo: Temporary failure in name resolution
Mar 17 18:01:31 perqin-nuc ss-local[485]:  2019-03-17 18:01:31 ERROR: failed to resolve the provided hostname
Mar 17 18:01:31 perqin-nuc systemd[1]: [email protected]: Main process exited, code=exited, status=255/EXCEPTION
Mar 17 18:01:31 perqin-nuc systemd[1]: [email protected]: Failed with result 'exit-code'.

Modifying /usr/lib/systemd/system/[email protected] and adding Wants after After works well:

[Unit]
Description=Shadowsocks-Libev Client Service
After=network-online.target
# Add the line below
Wants=network-online.target

[Service]
Type=simple
User=nobody
CapabilityBoundingSet=CAP_NET_BIND_SERVICE
ExecStart=/usr/bin/ss-local -c /etc/shadowsocks/%i.json

[Install]
WantedBy=multi-user.target

Tested on Arch Linux with latest shadowsocks-libev and NetworkManager packages installed.

[oakaigh]

@perqin If I remember correctly the “modern” version of shadowsocks-libev uses gethostbyname() (or getaddrinfo() as can be seen from the error message?) for domain resolution. So if it says “failure in domain resolution” it basically means either

• DNS server is unreachable (check internet connectivity)

OR

• the domain requested does NOT exist on servers specified in /etc/resolv.conf. (please double check the domain name you entered for spelling mistakes)

[perqin]

@PantherJohn It uses getaddrinfo(). The failure only happens during boot, and once I restart the service after having access to the terminal, it works properly. I am going to check the /etc/resolv.conf at failure some time later, but I think the issue is caused by no internet connection (because it starts before network-online.target is reached, as I mentioned before).