NOMAD_ALLOC_INDEX is 0 on all nodes when using system scheduler.

[siffiejoe]

Nomad version

Nomad v0.4.0

Operating system and Environment details

Linux host01 3.13.0-100-generic #147-Ubuntu SMP Tue Oct 18 16:48:51 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux

Issue

Two different nodes get the same allocation index (0) when using the system scheduler.

Reproduction steps

Create a nomad cluster of at least two nodes, copy the following script onto all nodes as /tmp/x.sh and run the job script below.

#!/bin/sh

while true; do
  echo "NOMAD_ALLOC_INDEX: $NOMAD_ALLOC_INDEX"
  sleep 1
done

All (both in my case) print the same allocation index (i.e. 0).

Job file (if appropriate)

job "xxx" {
    type = "system"

    datacenters = ["dc1"]

    task "x" {
    driver = "raw_exec"
    config {
        command = "/tmp/x.sh"
        args = []
    }
    resources {
        memory = 10
    }
    kill_timeout = 10s
    }
}

[dadgar]

Hey,

This is currently the desired behavior. The allocation index is used to indicate the instance out of a desired count size. System jobs do not have an explicit count but are based more on the size of a the targeted cluster, making the index less informative.

I'm curious what you have in mind for the index. Further there is an associated cost with making it a monotonically increasing counter (need to store additional meta data) and I am not sure of the benefit.

[siffiejoe]

I was trying to use the allocation index to get separate host names for each instance via consul. I need this because the service needs host names/addresses of all nodes that are going to be part of the cluster in advance (zookeeper is the culprit). I also want to wait until all instances are up before I start dependent services, and checking for zk-N.zookeeper.service.consul is an easy way to do that.

[jippi]

@siffiejoe I've been helping @smuthali with zookeeper as well, and he got something working out of that, perhaps drop by the gitter channel and give him a poke? :)

[dadgar]

@siffiejoe I think you could accomplish that by using node.unique.id or node.unique.name.

I am going to close this issue as it is desired behavior. Hope you get the help you need from @jippi and @smuthali

Best,
Alex

[jippi]

a hack is to have a docker entrypoint that echo ${NOMAD_ALLOC_INDEX} into the required file :)

[github-actions]

I'm going to lock this issue because it has been closed for 120 days ⏳. This helps our maintainers find and focus on the active issues.
If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.