Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

wreck: set nokz by default in rc1, support '-o kz' to force use of kvs for output #1896

Merged
merged 7 commits into from Dec 28, 2018

Conversation

Projects
None yet
3 participants
@grondo
Copy link
Contributor

grondo commented Dec 27, 2018

As discussed in #1865, add a -o kz option to more easily override global nokz setting.
Also add the option to flux-wreckrun(1) and flux-submit(1) and a simple test.

grondo added some commits Dec 27, 2018

wreck: support -o kz to override global nokz
Add new option `kz` which can be used to more simply override
a global `nokz` setting to force output for a single wreck job
to use kz streams in the kvs.
t2000-wreck: ensure -o kz overrides global nokz
Ensure that `-o kz` can be used in wreckrun to override global
'nokz' option.
doc: document kz option for flux-wreckrun(1),flux-submit(1)
Add documentation of `-o kz` to extra wreck options.
doc: add note about nokz to wreck operation section
Add a quick note about different output handling when -o nokz is used
to the flux-wreckrun(1) manpage.
doc/test: update spelling dictionary
Add "nokz" to spelling dictionary
wreck: fix nokz with --output/error on i686
Problem: nokz jobs using --output, --error fail on i686 becuase
of some kind of sign error when converting FLUX_NODEID_ANY from
C to Lua to json and back out again. Therefore, the check in
the output.lua plugin fails to detect that the io service should
be registered on rank 0 of the job.

Explicitly use -1 instead of FLUX_NODEID_ANY to work around this
issue for 32bit systems.

Fixes #1895
@codecov-io

This comment has been minimized.

Copy link

codecov-io commented Dec 27, 2018

Codecov Report

Merging #1896 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #1896      +/-   ##
==========================================
+ Coverage   80.13%   80.13%   +<.01%     
==========================================
  Files         196      196              
  Lines       35064    35064              
==========================================
+ Hits        28097    28099       +2     
+ Misses       6967     6965       -2
Impacted Files Coverage Δ
src/modules/wreck/wrexecd.c 76.74% <100%> (ø) ⬆️
src/modules/barrier/barrier.c 76.55% <0%> (-2.07%) ⬇️
src/cmd/flux-module.c 83.72% <0%> (-0.24%) ⬇️
src/modules/kvs/kvs.c 66.56% <0%> (-0.15%) ⬇️
src/common/libflux/message.c 81.64% <0%> (+0.12%) ⬆️
src/modules/connector-local/local.c 74.81% <0%> (+1.03%) ⬆️
@garlick

This comment has been minimized.

Copy link
Member

garlick commented Dec 27, 2018

Want to add a commit to make nokz the default in the installed rc1?

@grondo

This comment has been minimized.

Copy link
Contributor Author

grondo commented Dec 27, 2018

Ok, can do.

etc/rc1: make -o nokz default for all wreck jobs
Problem: Storing output streams for all jobs in the wreck prototype
has been show to cause excessive growth of the content store, and thus
is considered poor default behavior.

Add `flux wreck setopt nokz` to the default rc1 so that the "nokz"
operation for wreck jobs becomes the default behavior. This behavior
may be overidden with `-o kz` or `-o nokz=false` per job.

Fixes #1865

@grondo grondo changed the title wreck: support '-o kz' to force use of kvs for output wreck: set nokz by default in rc1, support '-o kz' to force use of kvs for output Dec 27, 2018

@grondo

This comment has been minimized.

Copy link
Contributor Author

grondo commented Dec 27, 2018

Ok, added flux wreck setopt nokz to rc1:

$ src/cmd/flux start -s 4 flux wreck getopt
flux-start: warning: setting --bootstrap=selfpmi due to --size option
nokz=true
@garlick

This comment has been minimized.

Copy link
Member

garlick commented Dec 27, 2018

Thanks! Restarted couple of hung builders.

@garlick

This comment has been minimized.

Copy link
Member

garlick commented Dec 28, 2018

Much appreciated!

@garlick garlick merged commit e9d9c02 into flux-framework:master Dec 28, 2018

1 check passed

continuous-integration/travis-ci/pr The Travis CI build passed
Details

@grondo grondo deleted the grondo:wreck-opt-kz branch Dec 28, 2018

@grondo

This comment has been minimized.

Copy link
Contributor Author

grondo commented Dec 28, 2018

Great thanks!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.