Specs:
Client version: 0.6.6
Go version (client): go1.2rc3
Git commit (client): 6d42040
Server version: 0.6.6
Git commit (server): 6d42040
Go version (server): go1.2rc3
Last stable version: 0.6.6
Linux mydockerhost 3.8.0-32-generic #47~precise1-Ubuntu SMP Wed Oct 2 16:19:35 UTC 2013 x86_64 x86_64 x86_64 GNU/Linux
Problem:
When trying to 'run' containers in parallel, about 1/2 of them fail.
Reproduction:
- Install docker 0.6.6
- Pull image of your choice.. I'm using 'centos'
- Create a script or in 2 separate terminals 'run' (using CLI, but REST API will repo as well) the image a handful or times in succession. In essence this is parallel 'run' of a single image and mimics what might happen if 2 users are trying to run an image at the same time.
- About 1/2 the run container calls fail.
Docker log snippet
[debug] api.go:999 Calling POST /containers/create
2013/11/07 09:19:25 POST /v1.6/containers/create
[debug] api.go:999 Calling POST /containers/{name:.*}/start
2013/11/07 09:19:25 POST /v1.6/containers/532ad60d936a/start
[error] api.go:1025 Error: Cannot start container 532ad60d936a: Cannot find child for /crimson_lion
[error] api.go:79 HTTP Error: statusCode=500 Cannot start container 532ad60d936a: Cannot find child for /crimson_
lion
[debug] api.go:999 Calling POST /containers/create
2013/11/07 09:19:25 POST /v1.6/containers/create
[error] api.go:1025 Error: Conflict, /crimson_lion already exists.
[error] api.go:79 HTTP Error: statusCode=409 Conflict, /crimson_lion already exists.
[debug] api.go:999 Calling POST /containers/{name:.*}/start
2013/11/07 09:19:25 POST /v1.6/containers/7bd5b4eb203c/start
[debug] container.go:1202 monitor: waiting for container 7bd5b4eb203c82721ec0ee0629e23a176b280312a2673182350c0ccf
6c7ce036 using cmd.Wait
On the CLI you'll see the same:
2013/11/07 09:19:25 Error: Cannot start container 532ad60d936a: Cannot find child for /crimson_lion
2013/11/07 09:19:25 Error: Conflict, /crimson_lion already exists.
Notes
- In the above log snippet I'm not using the -name parameter. However even if you generate a random name for each container this scenario still fails. Race condition?
Specs:
Problem:
When trying to 'run' containers in parallel, about 1/2 of them fail.
Reproduction:
Docker log snippet
On the CLI you'll see the same:
Notes