NIC Detection

Detection failure can be caused by a system error or a DHCP connection issue, in which the machine did not obtain the IP address from the DHCP server.

How to fix DHCP connection issue:

Maintenance – Check the Network, DHCP, and IP connectivity and make sure that everything is working properly.
Set a manual IP – Assign a manual IP address to the computer. This IP must be unique within the existing network (if it’s local, it should be unique within the local network).

If Swarm still does not locate any NICs, it displays the following message in the http://localhost:1113 Admin panel.

And in the Swarm log displays the following information:

[INFO - app] Running Swarm v2.0.15 7849af98
[WARN - app] Running in fallback mode [NO NICS]
. . . . . . .

In this case, set NICs manually via the swarm.yaml settings file. Currently, Swarm supports only IPv4.

  manualNics:
  - ip: 1.1.1.2
    name: mynic

A full specification is supported as well (MAC + IPv4 + IPv6 + name). However, a brief specification (ip + name) is sufficient: 

  manualNics:
  - mac: 11:12:13:14:15:16
    ip4: 1.1.1.44
    ip6: 2001:0000:130F:0000:0000:09C0:876A:130B
    name: mynic

Some NICs can be rejected, so Swarm ignores them. Swarm does not bind to a specific NIC while it communicates via TCP. By default, Swarm rejects the so-called loopback interface/NIC. Additionally, you can add your own rejection rules in the swarm.yaml settings file using the setting:

rejectNics: [<SOME-CRITERIA>]

Supported criteria are (examples): loopback, nocarrier, p2p, noarp, vm, vpn, tunnel, nogateway, nomac, nomulticast, name:eth12, ipv4:10.0.0.99. Use commas to separate the items in the list inside brackets. See this article for more information.

Some short notes: noarp - ARP is not set up, vm - NIC seems to belong to a virtual machine, nomac - MAC address is not set up, etc. For instance, name:eth12 will reject NIC with such name, and ipv4:10.0.0.99 will reject NIC with such IP (one of).

Some NICs have multiple IP addresses. Swarm recognizes only 1 IP by NIC, so it has an algorithm that chooses one IP only. By default, Swarm prefers IP marked as Preferred on Windows. Swarm chooses IP with the largest preferred lifetime (when it is available) and Swarm prefers the first IP when no other available criteria are available.
You can interfere with this: there is a special setting that can be added in the swarm.yaml settings file: nicIpPrio. It is a map of a list of priorities or sorting tactics (the key is the name of NIC), which are:

index – Sorts by index in the output of OS commands report.
preferred_lft – Sorts by “preferred lifetime” of IP.
unlikely <IPv4> – Marks the low IP priority. In case of multiple IPs, it is not chosen. Multiple unlikely priorities are allowed. 
likely <IPv4> – Marks this IP as a high priority for the specific NIC in case of multiple IPs. 

The key of nicIpPrio setting is a string name of the NIC or * if these priorities should be applied to all/any NIC. Two good empirical rules:

• Currently, the OS output command language is English. Local languages can be shown in Windows ipconfig. However, they are allowed only in the field values, so it will likely not break the parsing.

Workaround: Create a BAT file calling:

chcp 850
ipconfig /all

With a setting of the codepage using commands like chcp 437 or chcp 850 - see more here.

• Another possible issue is if OS commands used to determine NICs and their IPs cannot be found on the machine. If there are no errors, they were successfully detected programmatically. You can set the exact full path to OS commands in swarm.yaml settings file, as it follows: