cluster.routing.allocation.awareness.force does not work like promised

[toughcoding]

Elasticsearch Version

8.12.0

Installed Plugins

No response

Java Version

bundled

OS Version

Linux 1d1f1835b648 6.4.16-linuxkit #1 SMP PREEMPT Thu Nov 16 10:49:20 UTC 2023 aarch64 aarch64 aarch64 GNU/Linux

Problem Description

According to docs if not ALL nodes are available that will form complete list of cluster.routing.allocation.awareness.attributes
then replicas should not be STARTED but should be in UNASSIGNED state.

In practice it does not work. Not only when node_left cluster which I simulate on my blog and that can kill cluster but even when not all nodes that are part of cluster are being started.

Steps to Reproduce

Starting network, volumes and first node

docker network create elknodes
docker volume create --opt type=tmpfs --opt device=tmpfs --opt o=size=2m europe01data
docker volume create --opt type=tmpfs --opt device=tmpfs --opt o=size=2m africa01data

docker run --rm \
--name europe01 \
--net elknodes \
-d \
-e ES_JAVA_OPTS="-Xms2g -Xmx2g" \
-e node.name="europe01" \
-p 9200:9200 \
-e node.attr.continent="europe" \
-v europe01data:/usr/share/elasticsearch/data \
-e cluster.routing.allocation.awareness.attributes=continent \
-e cluster.routing.allocation.awareness.force.continent.values="europe,arctica,africa" \
docker.elastic.co/elasticsearch/elasticsearch:8.12.0

reset password

docker exec -it europe01 bash -c "(mkfifo pipe1); ( (elasticsearch-reset-password -u elastic -i < pipe1) & ( echo $'y\n123456\n123456' > pipe1) );sleep 5;rm pipe1"

get enrollment token

token=`docker exec -it europe01 elasticsearch-create-enrollment-token -s node | tr -d '\r\n'

Start second node, so values europe,africa are used but arctica is missing

docker run --rm \
-e ENROLLMENT_TOKEN="$token" \
-e node.name="africa01" \
-e node.attr.continent="africa" \
-v africa01data:/usr/share/elasticsearch/data \
-e cluster.routing.allocation.awareness.attributes=continent \
-e cluster.routing.allocation.awareness.force.continent.values="europe,arctica,africa" \
-p 9201:9200 \
--name africa01 \
--net elknodes \
-d \
-m 2GB docker.elastic.co/elasticsearch/elasticsearch:8.12.0

Create index

curl -k -u elastic:123456 -XPUT "https://localhost:9200/customerdata" \
-H 'content-type: application/json' -d'
{
  "settings": {
    "number_of_shards": 1,
    "number_of_replicas": 1
  }
}'

Check shard allocation

curl -k -u elastic:123456 -XGET "https://localhost:9200/_cat/shards?v&s=state:asc&index=customerdata"

will return

index        shard prirep state   docs store dataset ip         node
customerdata 0     r      STARTED    0  227b    227b 172.26.0.3 africa01
customerdata 0     p      STARTED    0  227b    227b 172.26.0.2 europe01

Planned node arctica was not started but Elasticsearch already assigned replicas.

Logs (if relevant)

No response

[DaveCTurner]

Thanks very much for your interest in Elasticsearch.

This appears to be a user question, and we'd like to direct these kinds of things to the Elasticsearch forum. If you can stop by there, we'd appreciate it. This allows us to use GitHub for verified bug reports, feature requests, and pull requests. Specifically, the behaviour you describe is expected.

There's an active community in the forum that should be able to help get an answer to your question. As such, I hope you don't mind that I close this.

[DaveCTurner]

Ah wait I see, the behaviour is as expected but the docs say something else. I'll fix that up in a sec. See #104800

[elasticsearchmachine]

Pinging @elastic/es-docs (Team:Docs)

[elasticsearchmachine]

Pinging @elastic/es-distributed (Team:Distributed)