introduce GetImageWithDetails to manage image inspect data in backend#27
introduce GetImageWithDetails to manage image inspect data in backend#27ndeloof merged 1 commit intorumpl:masterfrom
Conversation
ndeloof
force-pushed
the
image_inspect_ctrd
branch
from
3236817 to
3e175c0
Compare
image/image.go
Outdated
| Details *Details | ||
| } | ||
|
|
||
| // Details provides additional image data | ||
| type Details struct { | ||
| Size int64 | ||
| Metadata map[string]string | ||
| Driver string | ||
| LastUpdated time.Time |
There was a problem hiding this comment.
For the new fields, make sure they are not serialised to JSON (json:"-")
not sure if it works to add that to Details, or it if's needed for the individual fields (or both) (haven't tested)
There was a problem hiding this comment.
makes me wonder: why do we define json rules for an engine type? This one is not part of the API
There was a problem hiding this comment.
I think (but would have to double-check) these types are also used to serialise image information to disk (and as such could be part of, e.g., save/load. So in that case it's not for the API, but to prevent persisting this data to disk.
IIRC, this is also the reason why the V1Image is embedded, as that's what's used for save/load
ndeloof
force-pushed
the
image_inspect_ctrd
branch
from
fcdcb46 to
bb4c1bb
Compare
ndeloof
force-pushed
the
image_inspect_ctrd
branch
4 times, most recently
from
bb43aa3 to
fbb0432
Compare
| // Details provides additional image data | ||
| type Details struct { | ||
| Size int64 | ||
| Metadata map[string]string |
There was a problem hiding this comment.
Just realised that this is to store graphdriver metadata; wondering if that's clear enough from this type (at least could use a GoDoc), or was the intent to re-purpose this field for something else following the containerd integration?
There was a problem hiding this comment.
I actually wonder why graphdriver metadata are exposed to the API, seems like debugging data as a user would hardly use image's mergedir
There was a problem hiding this comment.
Yes, I agree. It looks like this was one of the Red Hat contributions, and I can only "guess" how they wanted to use this (possibly as part of some systemd hackery); moby#13198 (also related; moby#13130)
Signed-off-by: Nicolas De Loof <[email protected]>
ndeloof
force-pushed
the
image_inspect_ctrd
branch
3 times, most recently
from
2ef8a54 to
50401b4
Compare
3 participants
- What I did
Introduce GetImageWithDetails so that "image inspect" logic is not implement by router but backend, and we can re-implement using containerd snapshotter
relates to moby#43680