@samuelkarp I also want this feature, so I started to look at how it could be implemented. The other loggers use the function loggerutils.ParseLogTag in order to render the user provided template. Unfortunately this function will work only with the logger config option named tag, and will not check awslogs-stream.

Here is the relevant portion of the awslogs driver:
https://github.com/docker/docker/blob/692bfdaee3b48f4e5174bb80098692a0ded1aecf/daemon/logger/awslogs/cloudwatchlogs.go#L92

This shows how the syslog driver uses loggerutils.ParseLogTag:
https://github.com/docker/docker/blob/692bfdaee3b48f4e5174bb80098692a0ded1aecf/daemon/logger/syslog/syslog.go#L65

And here is the source for loggerutils.ParseLogTag:
https://github.com/docker/docker/blob/692bfdaee3b48f4e5174bb80098692a0ded1aecf/daemon/logger/loggerutils/log_tag.go#L14

Here are the docs for tag in the other drivers:
https://docs.docker.com/engine/reference/logging/log_tags/

What do you think a good way to handle this situation would be? Here are some ideas:

1. We can change the config option awslogs-stream to tag.
2. We can also allow a user to specify tag which will overwrite awslogs-stream.
3. We can rewrite loggerutils.ParseLogTag to allow specifying the option name to use.
4. We can write/duplicate the loggerutils.ParseLogTag code inside awslogs driver and hard code it to use the awslogs-stream.

What do you think?

@ianneub, thanks for pinging me.

Of your ideas, I'd lean toward number 3. Tags have a different meaning for many AWS resources, so I'd prefer to avoid introducing the potential for confusion.

+1

I'd also add the ability to use log tags in awslogs-group as well as awslogs-stream.

Are there any known workarounds for this limitation? Given awslog-streams should be unique per running container, I don't see any way to effectively override awslogs-stream.

ECS task definitions don't really have enough context to specify log groups / streams effectively (e.g. you likely want different log groups for different environments or stacks)

This one has been quiet for a while. I see a version/1.9 label applied, though I don't see any comments in that or any subsequent releases about this functionality. The documentation doesn't seem to indicate any different behavior for the awslogs driver and my test using the log-opt tag on version 1.10.1 is not supported with the awslogs driver. Does anyone know what the status is or future plans?

@CameronGo This would most likely have to be coded up by someone who wants this feature.
Right off the bat, I can say the syntax is not likely to be accepted, but I do believe we have support for go-template thin these options, and that would be the place to start.

@CameronGo the "1.9" label is automatically applied, based on the version an issue was reported with (in the top description)

+1 for this feature

@CameronGo @jhovell (any anyone else using ECS) We just added the ability to specify a stream prefix in the ECS task definition as awslogs-stream-prefix (the full stream name ends up being $prefix/$container-name/$task-id). See this blog post and our documentation for details.

@samuelkarp @CameronGo #27707 was merged, does that satisfy this feature request?

@thaJeztah Yes, I believe so.

Thanks @samuelkarp, I'll go ahead and close this issue, but @CameronGo let me know if there's still something that needs to be addressed 👍

Thanks @thaJeztah - yes that is working and fits my needs. Thanks for checking!

Good to hear @CameronGo!

Sorry, but this is not really fixed. Im using the prefix, and its alot better. thx!.

But i still got about 12 log streams, because they a grouped by the service ID.

DOMAIN/stage-DOMAIN/273ac3de-0bca-4bee-a35b-9e0206009380
DOMAIN/stage-DOMAIN/324c52f3-0f5c-4063-8b03-b490885aef3c
DOMAIN/stage-DOMAIN/355251f2-4e7f-4837-86c5-f2f187bf509c
DOMAIN/stage-DOMAIN/4cbd1975-3aad-46e5-a8bd-3241f2233bd0
DOMAIN/stage-DOMAIN/5cc3307b-c5ac-4f1c-9421-f2d2f1f9b0f6
DOMAIN/stage-DOMAIN/62b4cccf-c4a6-474e-8478-58912977168f
DOMAIN/stage-DOMAIN/71e60952-ce6d-4f8b-a901-445b811f6165
...

In an debugging situation, i need to click on all stream to find the error..

@ghost

could you share the docker run command and docker version?

I did the same as --log-opt awslogs-stream="$HOSTNAME-{{.ID}}", but I didn't get the docker container id properly.

docker run --log-driver=awslogs --log-opt awslogs-region=ap-southeast-2
  --log-opt awslogs-create-group=true
  --log-opt awslogs-group="hello-world"
  --log-opt awslogs-stream="$HOSTNAME-{{.ID}}"
  hello-world

Here is the log stream name I saw, it doesn't show container id of {{.ID}}

ip-10-0-2-123.ap-southeast-2.compute.internal-{{.ID}}

I did try these log_tags one by one, seems no one works.

https://docs.docker.com/config/containers/logging/log_tags/

{{.ID}} | The first 12 characters of the container ID.
{{.FullID}} | The full container ID.
{{.Name}} | The container name.
{{.ImageID}} | The first 12 characters of the container’s image ID.
{{.ImageFullID}} | The container’s full image ID.
{{.ImageName}} | The name of the image used by the container.
{{.DaemonName}} | The name of the docker program (docker).

Updates

I fixed to replace awslogs-stream with tag, they can't be used together.

  --log-opt tag="$HOSTNAME-{{.ID}}"