feat: try to fix default hostname

[NtskwK]

No description provided.

[pierotofy]

Hey thanks @NtskwK, I understand what the intent here is, but can you provide some context about how to reproduce the problem this patch is solving?

[NtskwK]

I noticed that the name of container are not always the same when I try to deploy the WebODM on WSL with ./webodm.sh start --dev --gpu(?) . Sometime, node-odm-1, webodm-node-odm-1, [DIR_NAME]-odm-1.

I don't fully understand the way nodeodm is named. But setting the hostname to the container name will allow them to connect successfully.

[pierotofy]

I wonder if explicitly setting the container_name (https://docker-docs.uclv.cu/compose/compose-file/compose-file-v2/#container_name) property to docker-compose.nodeodm.yml would make the hostname predictable and not require runtime checks.

If so, we could use it when --default-nodes is 1 in webodm.sh (but not when --default-nodes is > 1).

[NtskwK]

That's weird. If there is already a container named webodm_db/webodm_webapp/broker/worker, docker-compose will throw an error xxx is already exist then stop batch. But it's different to node-odm-1, because docker-compose.nodeodm.yml(and .gpu.*) dose not appoint container_name with container_name: xxx. So [DIR_NAME]-node-odm-1 will be renamed by Docker automaticlly.

[NtskwK]

> webodm.sh

down(){
	command="$docker_compose -f docker-compose.yml"

	if [ "${GPU_NVIDIA}" = true ]; then
		command+=" -f docker-compose.nodeodm.gpu.nvidia.yml"
	elif [ "${GPU_INTEL}" = true ]; then
		command+=" -f docker-compose.nodeodm.gpu.intel.yml"
	else
		command+=" -f docker-compose.nodeodm.yml"
	fi

	command+=" -f docker-compose.nodemicmac.yml down --remove-orphans"

	run "${command}"
}

docker-compose.nodeodm.gpu.* will not be shutdown if --gpu was not be used in ./webodm down. This can be an important cause of container residue. Then, a new WebODM program starter will launch a container which is named DIR_NAME]-node-odm-1.

Use container name to replace hostname is only functional when both they(Client/Server) are in the same docker-compose group and can't be captured by other container or even by host machine. Well, the new WebODM cannot capture the already existing nodes(node-odm-1)

[NtskwK]

On user-defined networks, containers can not only communicate by IP address, but can also resolve a container name to an IP address.

Is can only works in user-defined bridge networks and disable in default bridge networks.

[NtskwK]

I wonder if explicitly setting the container_name (https://docker-docs.uclv.cu/compose/compose-file/compose-file-v2/#container_name) property to docker-compose.nodeodm.yml would make the hostname predictable and not require runtime checks.

If so, we could use it when --default-nodes is 1 in webodm.sh (but not when --default-nodes is > 1).

It's theoretically possible, but not recommended. The logic of the function will become complex. If docker-compose starts successfully, but docker-compose.node is blocked due to duplicate names, it will be deployed in a state without nodes, which is not expected.

[pierotofy]

https://docs.docker.com/compose/networking/

Mm, perhaps just setting COMPOSE_PROJECT_NAME to "webodm" would do?

[NtskwK]

Of course that's okay. Please turn it into a draft. I will try to modify the webodm.sh script in passing.

…

________________________________ 发件人: Piero Toffanin ***@***.***> 发送时间: 2024年9月5日 0:48 收件人: OpenDroneMap/WebODM ***@***.***> 抄送: NtskwK ***@***.***>; Mention ***@***.***> 主题: Re: [OpenDroneMap/WebODM] feat: try to fix default hostname (PR #1550) https://docs.docker.com/compose/networking/ Mm, perhaps just setting COMPOSE_PROJECT_NAME to "webodm" would do? ― Reply to this email directly, view it on GitHub<#1550 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AVDTMEBOGPEWMJBDKGSRBUDZU42VHAVCNFSM6AAAAABNRWUUTGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGMRZGU2DQMBYGE>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[NtskwK]

#1237

[pierotofy]

I'm not sure I understand the change here.

[NtskwK]

The shutdown function now has to be entered after the --gpu to shut down the gpu node. But it's just the startup parameter in the manual. There may be users who get unexpected results. I would like to be able to make sure they shut down.

[pierotofy]

Did you test with:

./webodm.sh restart --default-nodes 3 ?

[NtskwK]

My apologies. I forgot there would be conflict here.

I still don't fully understand the way to the naming. Can you give me some advice?

[pierotofy]

You need to understand how docker-compose works.

This is a good start: https://docs.docker.com/compose/

Read it all.

[NtskwK]

You are right, maybe I should not affect the behavior of the other. Just focus on the try to fix default hostname like next function do.

Workaround for mysterious "webodm_node-odm-1" or "webodm-node-odm-1" hostname switcharoo on Mac

How do you think？