Initial Travis configuration (rebased and fixed)#23
Conversation
* docker-compose is run in ci/sge so volume needs to be relative to ci/sge * run pytest in docker container * flake8 tweaks
|
Full disclosure: I am certainly not a docker expert, so I did the minimal change to get Travis to work without guarantee that this follows best docker practices. |
|
I'm very glad to see you working on this @lesteve :) I'm curious have you been using dask-jobqueue on your own? Also, to reduce build times we might want to add the |
|
@lesteve - this makes me very happy to see. Thanks for stepping up. No strong objections from me on the changes you made. |
|
It works! That's quite exciting :) Now we just need to make an SGE system, which should be comparatively easy :) |
|
@jhamman now that we have a system in place do you think some of the system administrators at NCAR would have time an interest in submitting a solution for PBS? |
Happy this is useful! I have been watching dask-drmaa and dask-jobqueue lately with the not-so-far-in-the-future goal to make our local SGE cluster more user-friendly for researchers. I have used only dask-drmaa so far and only for toy problems. |
|
Sounds good. If we can gather enough bug reports and usage issues in the
next couple weeks then maybe this is something we can work on when I'm in
town.
…On Tue, Mar 27, 2018 at 10:09 AM, Loïc Estève ***@***.***> wrote:
I'm very glad to see you working on this @lesteve
<https://github.com/lesteve> :) I'm curious have you been using
dask-jobqueue on your own?
Happy this is useful! I have been watching dask-drmaa and dask-jobqueue
lately with the near-future goal to make using our local SGE cluster more
user-friendly for researchers. I have used only dask-drmaa so far and only
for toy problems.
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#23 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AASszE9PbVqdJvqB-t_xWFKGEUtal6Adks5tikgpgaJpZM4S881t>
.
|
|
It would be great, I definitely plan to try test dask on our SGE cluster more extensively by then! |
|
I am also curious to see if adaptive scaling is effective for you. You
might want to try the adapt method of the cluster classes.
…On Tue, Mar 27, 2018 at 11:25 AM, Loïc Estève ***@***.***> wrote:
It would be great, I definitely plan to try test dask on our SGE cluster
more extensively by then!
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#23 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AASszEyw6V0vva4FdQPvQ6RPH5CZUpKQks5tilnSgaJpZM4S881t>
.
|
|
The comment about NCAR is probably optimistic. I would not assume that the
system administrators there will have time to help with this.
…On Tue, Mar 27, 2018 at 11:29 AM, Guillaume EB ***@***.***> wrote:
@mrocklin <https://github.com/mrocklin> @jhamman
<https://github.com/jhamman>, I was thinking about working on a
comparable solution for PBS. But if someone else is willing to work on this
that's fine with me.
Maybe we should open a specific issue for PBS, and same with Slurm?
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#23 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AASszFJj-4T5hWh7ePp292JYbzXmWvGpks5tilrDgaJpZM4S881t>
.
|
|
@jhamman ok to merge? |
4 participants
I fixed the conflicts in #13 and got Travis working on my fork, see this Travis build.
Closes #13.
I would have opened a PR on @jhamman fork but because of the conflicts this is not very convenient so I opened a new PR instead.