matsengrp / cft

Clonal family tree
5 stars 3 forks source link

How to deploy cft web server? #20

Closed cswarth closed 5 years ago

cswarth commented 7 years ago

Placeholder for information on how deploy cft web server.

cswarth commented 7 years ago

Dirk responded,

Chris, you can use this mechanism, you can install your app inside docker or ignore docker and use the main vm. https://teams.fhcrc.org/sites/citwiki/SciComp/Pages/How%20to%20run%20your%20own%20Docker%20host.aspx Dirk

I made and destroyed a VM using this method yesterday but have not populated it with our code and started a server yet. Given the complexities of establishing the environment, a Dockerfile is probably the best way to go. I don't yet know what filesystems are available to these VMs but they do not seem to have the full complement of /fh/fast mount points.

cswarth commented 7 years ago

i wrote to scicomp again, asking how to access our data from VMs created from their proxmox platform,

Is there a way for the new machine to access data that is elsewhere on the Hutch network? Could I mount remote filesystems? I'd need Hutch network ids and authentication to do that, right?

We will be running an internal-only web service and potentially have a fair amount of data that will be preprocessed elsewhere. I'm hoping there is a another answer besides rsync or scp.

Not via NFS. There's rsync and scp as you mention, http if you want or can use the object store (econofile). The remaining option for a filesystem mount would be CIFS. I'm not real clear on what the application will be doing- what file systems would you need access to?

matsen commented 7 years ago

Well, it looks like the shortest path is to use scp, and if they complain that we're using too much data perhaps they will provide a more attractive option.

cswarth commented 7 years ago

Needs polish but cft web service is deployed on http://cftweb4.fhcrc.org:5000/ Just some sample trees are up there now. Tomorrow I'll work on getting more data and making it easier to deploy.

matsen commented 7 years ago

Man, we are going to miss you, @cswarth 😭

cswarth commented 7 years ago

I've tried to establish a larger data set for cftweb to serve up on Tuesday but have been thwarted by #48 and #49. Once these are fixed someone can rerun bin/postpartis and then run scons to complete the pipeline. The resulting output directory should be copied to cftweb4.fhcrc.org:~cftweb/output, username cftweb, passwd cftweb, i.e.

scp -r output cftweb4.fhcrc.org:~cftweb/

Then restart the cftweb server with sudo systemctl restart docker-cftweb-server.service, i.e.

ssh cftweb@cftweb4.fhcrc.org  sudo systemctl restart docker-cftweb-server.service

Results will be visible on http://cftweb4.fhcrc.org:5000

metasoarous commented 7 years ago

We're now looking with Dirk at waiting a little bit for a new container management system called Rancher that is being worked on. This would deploy onto the Hutch's AWS VPC. There are still some details there around how internal resources like storage are going to be accessed, but they're hoping to have something we can start testing by some time in January. There is also a Wercker plugin we can take a look at here (https://github.com/nhumrich/step-deploy-to-rancher) for CI.

For now, we're going to limp by maintaining a silly deployment to stoat, since I still don't have prox working yet :-(