core dump, docker official image, export ERL_FLAGS=" -name couchdb "

[juanbrein]

Description

When I run the official docker image using these environment variables it cores dump:

export COUCHDB_ERLANG_COOKIE=XXX
export COUCHDB_PASSWORD=XXX
export COUCHDB_SECRET=XXX
export COUCHDB_USER=admin
export ERL_FLAGS=" -name couchdb "

Steps to Reproduce

$ docker run -ti --rm \
--env COUCHDB_ERLANG_COOKIE=JIxa5jYokaO2IlE5Hf1l \
--env COUCHDB_PASSWORD=password \
--env COUCHDB_SECRET=77qz5K82uAkbIrBFBsgG \
--env COUCHDB_USER=admin \
--env COUCHDB_VERSION=3.3.3 \
--env ERL_FLAGS=" -name couchdb " \
couchdb
{"Kernel pid terminated",application_controller,"{application_start_failure,kernel,{{shutdown,{failed_to_start_child,net_sup,{shutdown,{failed_to_start_child,net_kernel,{'EXIT',nodistribution}}}}},{kernel,start,[normal,[]]}}}"}
Kernel pid terminated (application_controller) ({application_start_failure,kernel,{{shutdown,{failed_to_start_child,net_sup,{shutdown,{failed_to_start_child,net_kernel,{'EXIT',nodistribution}}}}},{kernel,start,[normal,[]]}}})

Crash dump is being written to: erl_crash.dump...done

Expected Behaviour

Should be able to run couchdb

Your Environment

• CouchDB version used: 3.3.3
• Operating system and version: ubuntu:22.04

Additional Context

The reason I need this is because I run couchdb in Kubernetes using the official helm chart. I also have a plain docker compose configuration in my local machine. When I copy the data from one environment to another I get the infamous error message "This database failed to load."

I read on different forums that the cluster name has to be the same in order for that work.

I can't find any documentation related unfortunately.

[nickva]

Thank you for your report @juanbrein

{'EXIT',nodistribution}} indicates an issue with erlang distribution (that's the erlang VM inter-node clustering system).

Unfortunately I don't use couchdb with docker much so don't have an exact reason for the failure. It may be a networking issue or a permissions issue.

Also try to take a look at https://hub.docker.com/_/couchdb it's one of experiments of creating a cluster with compose and running some basic load testing https://gist.github.com/nickva/cf9cac975ceb062b61872f7857a0e104. Maybe that will give you some hints.

[juanbrein]

Thanks for the hint! I tried using NODENAME, but the problem is that it sets the -name flag to couchdb@127.0.0.1 if NODENAME is 127.0.0.1 ... and that name is different from the one assigned on kubernetes -name couchdb ....

Another way would be to change the node name in the shard files... but that is not documented either...

I have to say that even though the restore process from / to same configurations work... if you are in scenarios like myself... which I believe is not so radical... it doesn't work, the errors are not very useful and is not documented...

[big-r81]

Thanks for the hint! I tried using NODENAME, but the problem is that it sets the -name flag to couchdb@127.0.0.1 if NODENAME is 127.0.0.1 ... and that name is different from the one assigned on kubernetes -name couchdb ....

This comes from here. So you get two different -name ... settings, if you set it in ERl_FLAGS too.