[feat] Allow flexible node allocation to select nodes in any state

[teojgo]

Fixes #1032

[vkarak]

This PR is mixed with commits from other PRs. Can you fix it?

[teojgo]

This PR is mixed with commits from other PRs. Can you fix it?

Yes I made a mistake when pushing because it have forked from another branch.

[pep8speaks]

Hello @teojgo, Thank you for updating!

Cheers! There are no PEP8 issues in this Pull Request!Do see the ReFrame Coding Style Guide

Comment last updated at 2020-06-22 08:32:08 UTC

[codecov-commenter]

Codecov Report

Merging #1374 into master will increase coverage by 0.06%.
The diff coverage is 95.65%.

@@            Coverage Diff             @@
##           master    #1374      +/-   ##
==========================================
+ Coverage   91.68%   91.75%   +0.06%     
==========================================
  Files          83       83              
  Lines       12714    12712       -2     
==========================================
+ Hits        11657    11664       +7     
+ Misses       1057     1048       -9     

Impacted Files	Coverage Δ
reframe/frontend/cli.py	78.15% <ø> (+0.17%)	⬆️
unittests/test_schedulers.py	93.82% <93.75%> (+1.59%)	⬆️
reframe/core/schedulers/__init__.py	98.83% <100.00%> (ø)
reframe/core/schedulers/local.py	100.00% <100.00%> (ø)
reframe/core/schedulers/slurm.py	57.54% <100.00%> (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 8bf8e1e...03c923c. Read the comment docs.

[ekouts]

lgtm

[vkarak]

I can still see stale changes from the constraint PR. Plus there is another error that exists also in master most probably. If you pass --flex-alloc-nodes without an argument, it will complain, whereas it should just assume idle.

[teojgo]

I can still see stale changes from the constraint PR. Plus there is another error that exists also in master most probably. If you pass --flex-alloc-nodes without an argument, it will complain, whereas it should just assume idle.

But isn't the point of having it to pass an argument. There is no default if the argument is used. Should I put one?

[vkarak]

@teojgo I think we say that the default is idle and here we imply that passing an argument is optional, which is not the case as it is now.

[teojgo]

Yeah not using the option at all implies idle

[teojgo]

But if you use the option an argument is required

[vkarak]

Ok, now I get it. But we should still change the documentation of the command line option:

--flex-alloc-nodes=POLICY
blah blah blah

If this option is not specified, the default allocation policy for flexible tests is 'idle'.

[vkarak]

lgtm