[LIBCLOUD-954] ec2: Allow cn-north-1 even without pricing

[pquentin]

ec2: Allow cn-north-1 even without pricing

Description

See https://issues.apache.org/jira/browse/LIBCLOUD-954 and 1e1d77f

@fjros and @duanshiqiang, what do you think?

Status

• done, ready for review

Checklist (tick everything that applies)

• Code linting (required, can be done after the PR checks)
• Documentation
• Tests
• ICLA (required for bigger changes)

[fjros]

Thanks @pquentin, a nice trade-off :-)

Maybe this should be clearly stated in the docs, since in practice there's at least one api method (I think) not supported for the region.

[duanshiqiang]

I think this is good :)

[pquentin]

@fjros Thanks. I tried a better approach: setting price to None when it's not available as with cn-north-1. Also improved the docs. Please tell me what you think. I highly appreciate your timely feedback!

[codecov-io]

Codecov Report

Merging #1127 into trunk will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##            trunk    #1127      +/-   ##
==========================================
+ Coverage   85.45%   85.45%   +<.01%     
==========================================
  Files         346      346              
  Lines       66272    66278       +6     
  Branches     5899     5900       +1     
==========================================
+ Hits        56631    56637       +6     
  Misses       7239     7239              
  Partials     2402     2402

Impacted Files	Coverage Δ
libcloud/compute/drivers/ec2.py	74.06% <100%> (+0.05%)	⬆️
libcloud/test/compute/test_ec2.py	97.91% <100%> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update a1c0e60...65fb59b. Read the comment docs.

[fjros]

@pquentin Great!! As far as I'm concerned, this fixes the issue.

Just a minor note... Maybe _get_size_price() should directly return None instead of raising an exception. In that way, the behavior would be consistent across the codebase. I mean, just moving the try..except block from list_sizes() to _get_size_price(). What do you think?

[pquentin]

Ah, yes, that would be better! I'll do that.

[pquentin]

Thinking a bit more about it, that would affect other drivers. I would not want to return None silently in those cases because we would not be alerted of an issue in tests. The change as it is currently only affects EC2.

[pquentin]

Merged in trunk.

@fjros Please fell free to open a new PR or JIRA issue if you think of a better way. Thanks again.
@duanshiqiang Can you please close your JIRA issue if that fixes your problem? Thanks.