[arvados] about slurm job using through container API

Yongjian Guo guoyj2003 at gmail.com
Sun Sep 10 15:10:57 EDT 2017


I have solved my problem. One of my keepstore is configured wrongly. I
should use the internal IP address instead the hostname.

I have one more question about the speed of loading Docker image on compute
node. How long do we expect to load in a 500M images from the keep storage?
I feel the loading speed of mine is pretty slow. But I do not know the
benchmark.

Also, once a docker image is loaded from keep? Will it be kept in cache for
a while? If so, how long is it kept in the cache?

Thanks,

Yongjian



On Sat, Sep 9, 2017 at 11:02 PM, Yongjian Guo <guoyj2003 at gmail.com> wrote:

> Hi, Everyone,
>
> When I submit a job through the container API, I saw the the following
> output on the compute node:
>
> zzzzz-dz642-18tx6hpy1j48oz2 2017-09-10T02:45:36.883643677Z Executing
> container 'zzzzz-dz642-18tx6hpy1j48oz2'
> zzzzz-dz642-18tx6hpy1j48oz2 2017-09-10T02:45:36.883706265Z Executing on
> host 'compute8'
> zzzzz-dz642-18tx6hpy1j48oz2 2017-09-10T02:45:36.903586911Z Fetching Docker
> image from collection '62b8bd1d1a65c04571d0aece47172b07+342'
> zzzzz-dz642-18tx6hpy1j48oz2 2017-09-10T02:45:36.919914767Z Using Docker
> image id 'sha256:6c4e1452f1a36c2b8255fba7097147
> aff739f98cd05a8c82f9214df4d3dc801d'
> zzzzz-dz642-18tx6hpy1j48oz2 2017-09-10T02:45:36.920946105Z Loading Docker
> image from keep
> zzzzz-dz642-18tx6hpy1j48oz2 2017-09-10T02:45:52.357768811Z While loading
> container image: While loading container image into Docker: error during
> connect: Post http://%2Fvar%2Frun%2Fdocker.sock/v1.21/images/load?quiet=0:
> unexpected EOF
> zzzzz-dz642-18tx6hpy1j48oz2 2017-09-10T02:45:52.357810953Z Cancelled
>
> Looks like the computer node is trying to load the docker image
> (arvados/job) but is not successful. The configuration on my compute node
> must be wrong. Could you give me more information on how to fix this
> problem?
>
> Also, is it trying to load the image from keep? How can I check if the
> keep storage has been mounted?
>
> Thanks a lot.
>
> Jason
>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.arvados.org/pipermail/arvados/attachments/20170910/47e73eaa/attachment.html>


More information about the arvados mailing list