addprocs(1) fails when some ethernet interfaces report the speed and others not

[ufechner7]

addprocs(1) fails because there's a bug in the DistributedNext package version 1.0.0. The error occurs in the init_bind_addr() function where it tries to call isless(::Float64, ::Nothing), which is not a valid comparison.

MWE:

julia> using DistributedNext
julia> addprocs(1)

Output:

ERROR: MethodError: no method matching isless(::Float64, ::Nothing)
The function `isless` exists, but no method is defined for this combination of argument types.

Closest candidates are:
  isless(::Missing, ::Any)
   @ Base missing.jl:87
  isless(::Any, ::Missing)
   @ Base missing.jl:88
  isless(::T, ::T) where T<:Union{Float16, Float32, Float64}
   @ Base float.jl:630
  ...

Stacktrace:
  [1] _rf_findmax(::Tuple{Float64, Int64}, ::Tuple{Nothing, Int64})
    @ Base ./reduce.jl:907
  [2] BottomRF
    @ ./reduce.jl:86 [inlined]
  [3] MappingRF
    @ ./reduce.jl:100 [inlined]
  [4] _foldl_impl(op::Base.MappingRF{…}, init::Base._InitialValue, itr::Base.Pairs{…})
    @ Base ./reduce.jl:62
  [5] foldl_impl
    @ ./reduce.jl:48 [inlined]
  [6] mapfoldl_impl
    @ ./reduce.jl:44 [inlined]
  [7] mapfoldl
    @ ./reduce.jl:175 [inlined]
  [8] _findmax
    @ ./reduce.jl:906 [inlined]
  [9] findmax
    @ ./reducedim.jl:1198 [inlined]
 [10] init_bind_addr()
    @ DistributedNext ~/.julia/packages/DistributedNext/Ujy0m/src/cluster.jl:1334
 [11] init_multi()
    @ DistributedNext ~/.julia/packages/DistributedNext/Ujy0m/src/cluster.jl:1353
 [12] addprocs(manager::DistributedNext.LocalManager; kwargs::@Kwargs{})
    @ DistributedNext ~/.julia/packages/DistributedNext/Ujy0m/src/cluster.jl:460
 [13] addprocs
    @ ~/.julia/packages/DistributedNext/Ujy0m/src/cluster.jl:459 [inlined]
 [14] #addprocs#279
    @ ~/.julia/packages/DistributedNext/Ujy0m/src/managers.jl:490 [inlined]
 [15] addprocs(np::Int64)
    @ DistributedNext ~/.julia/packages/DistributedNext/Ujy0m/src/managers.jl:487
 [16] top-level scope
    @ REPL[3]:1
Some type information was truncated. Use `show(err)` to see complete types.

Output of versioninfo()

julia> versioninfo()
Julia Version 1.11.6
Commit 9615af0f269 (2025-07-09 12:58 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 32 × AMD Ryzen 9 7950X 16-Core Processor
  WORD_SIZE: 64
  LLVM: libLLVM-16.0.6 (ORCJIT, znver4)
Threads: 1 default, 0 interactive, 1 GC (on 32 virtual cores)
Environment:
  LD_LIBRARY_PATH = /lib:/usr/lib:/usr/local/lib

This is on Ubuntu 22.04.

[giordano]

You'll need to elaborate what system you're using, since the code depends a lot on your network. Reading the code

DistributedNext.jl/src/cluster.jl

Lines 1328 to 1334 in c68111a

	# Pick the interface with the highest negotiated speed, if any
	interfaces_with_speed = filter(x -> !isnothing(x.speed), interfaces)
	if isempty(interfaces_with_speed)
	# If none of them report speed just pick the first one
	bind_addr = string(interfaces[1].ip)
	else
	idx = findmax(x -> x.speed, interfaces)[2]

my guess is that you have some interfaces which report the speed and some which don't, the code seems to handle only the cases where either the speed is always reported or never. What's the output of

using Sockets, DistributedNext
let
    interfaces = DistributedNext._get_interfaces(IPv4)
    if isempty(interfaces)
        # Include IPv6 interfaces if there are no IPv4 ones
        interfaces = DistributedNext._get_interfaces()
    end
    interfaces
end

?

@JamesWrigley probabbly

DistributedNext.jl/src/cluster.jl

Line 1334 in c68111a

idx = findmax(x -> x.speed, interfaces)[2]

should use interfaces_with_speed since that's otherwise unused?

[ufechner7]

2-element Vector{DistributedNext.Interface}:
 DistributedNext.Interface("enp12s0", :v4, ip"192.168.8.114", DistributedNext.ARPHardware_Ethernet, 2500.0)
 DistributedNext.Interface("wlp13s0", :v4, ip"192.168.1.170", DistributedNext.ARPHardware_Ethernet, nothing)

[JamesWrigley]

Thanks for the report, could you try #29 and see if that fixes it?