Closed aw32 closed 4 months ago
Sorry for the late reply.
Do you mean to get the job archive as a regular user or as admin getting the complete archive? The job archive can get pretty big, and I guess it is not feasible to download usin a API.
We plan to provide a S3 based backend for the job archive. This would then allow to access or download the job archive from a network location independent of cc-backend. Would this be a solution to your problem?
Do you mean to get the job archive as a regular user or as admin getting the complete archive? The job archive can get pretty big, and I guess it is not feasible to download usin a API.
There is already a REST API that is accessible with an API token. With "complete archive" I mean the JobArchive files (meta.json and data.json) for one job as it is stored on disk. Right now the REST API only allows to get certain metrics for a job, but that is quite cumbersome.
We plan to provide a S3 based backend for the job archive. This would then allow to access or download the job archive from a network location independent of cc-backend. Would this be a solution to your problem?
We already can access the files by using an additional method (e.g. network filesystem, ssh, etc.). This issue is about extending the REST API with a simple and easy way.
Ok, sorry for the misunderstanding. What we could do to send the complete information that is contained in both meta.json and data.json as one large json (with keys meta and data). Or do you need the files specifically?
Another option would be a flag to also include all metrics and extend the current job schema with the metric data. What do you think? Or do you have another suggestion?
I think the first suggestion (one large json with data and meta key) would be the easiest. If necessary, the json can be easily split into the two original json files.
It would also be useful to get the cluster specifications for the available clusters. In case of a change, one could easily get the current one using the API. Edit: To be more specific: this could be a separate API call that simply returns an array with all the cluster specifications.
Ok, that sound doable. I see if we can do it for the next release.
At the moment one needs to do multiple API calls to reconstruct the information of the JobArchive or get the JobArchive from the file system of the backend server. It would be nice to have one API call to receive the complete JobArchive.