Comment out spammy log message during image pull

[dperny]

Commented out the spammy log message during image pull. The existing
code seems adequate to report most errors. There may be a more correct
way to report pull status, but I don't know what it is.

Closes #935

Signed-off-by: Drew Erny [email protected]

[dperny]

@stevvooe @aaronlehmann PTAL. Nothing suggests to me that simply commenting out the offending log message isn't a passable solution, but I might be overlooking something.

[dperny]

whoops this is what i get for not building locally one a 1-line change.

[aaronlehmann]

I think it's okay to remove the pull progress for now, until we have a better way of reporting it.

[codecov-io]

Current coverage is 53.95%

Merging #1061 into master will decrease coverage by 0.03%

@@             master      #1061   diff @@
==========================================
  Files            75         75          
  Lines         11686      11681     -5   
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
- Hits           6310       6303     -7   
- Misses         4477       4479     +2   
  Partials        899        899          

Powered by Codecov. Last updated by 5667c2d...11d146c

[stevvooe]

I'd prefer we fix the todo... I leave in spammy messages for a reason. When the architecture is correct, this message goes away.

[dperny]

I don't understand what the correct architecture should be. What is important about reporting progress? I don't understand the problem well enough to fix it.

[aluzzardi]

Ping @stevvooe: could you address @dperny's question so we can move this forward?

[stevvooe]

I don't understand what the correct architecture should be. What is important about reporting progress? I don't understand the problem well enough to fix it.

We need to provide visibility into the pull process.

We can probably report this through container status. I'd like to see the changes proposed in docker/engine-api#89 (comment) and we can work from there.

We may also want to flow some of this information into the message field on TaskStatus.

[aluzzardi]

I don't think we can use container status - it would put too much pressure into the store.

Actually, we're already sending too many updates - the agent eventually should probably compact those reports.

Writing to raft is about the most expensive thing we can do in the system.

[stevvooe]

Writing to raft is about the most expensive thing we can do in the system.

Agreed, but we need some visibility into pull.

the agent eventually should probably compact those reports.

The protocol already supports this. We just need to make the agent do it.