Container secret mounts » History » Version 5
Peter Amstutz, 03/01/2018 04:06 PM
1 | 1 | Tom Clegg | h1. Container secret mounts |
---|---|---|---|
2 | |||
3 | 2 | Tom Clegg | This is a proposed modification to [[Containers API]]. |
4 | |||
5 | Add a "secret_mounts" serialized field to containers and container requests. |
||
6 | |||
7 | "secret_mounts" has the same form and behavior as "mounts", except: |
||
8 | * Only literal content is allowed (kind=text or kind=json) |
||
9 | * Current value is never returned in a container request or container API response |
||
10 | * Current value can be retrieved from a new API (@/arvados/v1/containers/$uuid/secret_mounts@) which must be authenticated by the container's own runtime token |
||
11 | 1 | Tom Clegg | * Never appears in container logs |
12 | * Never appears in the Arvados logs table |
||
13 | * Never appears in websocket updates |
||
14 | * Never appears in API server request logs |
||
15 | |||
16 | 2 | Tom Clegg | It is an error for the same key (mount path) to appear in both mounts and secret_mounts. It should not be possible to _commit_ a container request with this error condition, although it might be possible to _save._ |
17 | |||
18 | A secret_mounts key (path) cannot be |
||
19 | * equal to the container's output path, |
||
20 | * a descendant of the container's output path, or |
||
21 | * an ancestor of the container's output path (currently, this is a moot point because secret mounts are not directories) |
||
22 | |||
23 | 4 | Peter Amstutz | (PA) I would modify this restriction. Secret file mounts should be allowed to appear under the output directory, but must be _excluded_ when capturing output. The reasoning is that this aligns better with the CWL InitialWorkDir feature which pre-populates the output directory. Pre-populating other directories is technically legal but less portable (only works if you are using containers, which is an ok assumption for Arvados but not for CWL generally). |
24 | 3 | Peter Amstutz | |
25 | 2 | Tom Clegg | For clarity, some ways in which secret_mounts behaves like mounts are: |
26 | * Non-identical secret_mounts disqualifies a container for reuse. The mere existence of secret_mounts does not disqualify. |
||
27 | * secret_mounts can be set via container_requests#create and container_requests#update APIs |
||
28 | * secret_mounts cannot be null, but can be an empty hash |
||
29 | * keys of secret_mounts are paths in the container's filesystem, and always begin with "/" |
||
30 | 5 | Peter Amstutz | |
31 | h3. Secret environment |
||
32 | |||
33 | Add a "secret_environment" serialized field to containers and container requests. |
||
34 | |||
35 | "secret_environment" has the same form and behavior as "environment", except: |
||
36 | |||
37 | * Current value is never returned in a container request or container API response |
||
38 | * Current value can be retrieved from a new API (@/arvados/v1/containers/$uuid/secret_environment@) which must be authenticated by the container's own runtime token |
||
39 | * Never appears in container logs |
||
40 | * Never appears in the Arvados logs table |
||
41 | * Never appears in websocket updates |
||
42 | * Never appears in API server request logs |
||
43 | |||
44 | It is an error for the same key (environment var) to appear in both environment and secret_environment. It should not be possible to _commit_ a container request with this error condition, although it might be possible to _save._ |
||
45 | |||
46 | For clarity, some ways in which secret_environment behaves like environment are: |
||
47 | * Non-identical secret_environment disqualifies a container for reuse. The mere existence of secret_environment does not disqualify. |
||
48 | * secret_environment can be set via container_requests#create and container_requests#update APIs |
||
49 | * secret_environment cannot be null, but can be an empty hash |
||
50 | * keys of secret_environment are names of environment variables |
||
51 | |||
52 | h3. Secret command |
||
53 | |||
54 | |||
55 | Add a "secret_environment" serialized field to containers and container requests. |
||
56 | |||
57 | "secret_command" has the same form and behavior as "command", except: |
||
58 | |||
59 | * Current value is never returned in a container request or container API response |
||
60 | * Current value can be retrieved from a new API (@/arvados/v1/containers/$uuid/secret_command@) which must be authenticated by the container's own runtime token |
||
61 | * Never appears in container logs |
||
62 | * Never appears in the Arvados logs table |
||
63 | * Never appears in websocket updates |
||
64 | * Never appears in API server request logs |
||
65 | |||
66 | To run, "secret_command" is merged with "command". If secret_command is empty, it is ignored, otherwise they must be arrays of the same size. Entries in "command" which are "null" take their value from the corresponding entry in "secret_command". It is an error for the same entry in command and secret_command to to both have a value or both be null. |
||
67 | |||
68 | For clarity, some ways in which secret_command behaves like command are: |
||
69 | * Non-identical secret_command disqualifies a container for reuse. The mere existence of secret_command does not disqualify. |
||
70 | * secret_command can be set via container_requests#create and container_requests#update APIs |
||
71 | * secret_command cannot be null, but can be an empty list |