Helm Installation Queries

[YasirKusay]

Hello,

I am interesting in doing an on prem installation of Seqr, using the helm method primarily due to the many new features it provides, such as the ability for users to upload their own data, SV and long read support, etc. While the installation process appears relatively simple, there are a few things that are not clear from the guide:

• Once you finish up the installation steps, I am assuming that you can just use a browser and navigate to your own device's IP and can use the app there.
• There appears to only be http support at the moment correct? Is there a method to set it up with https and our own domain (provided we fully acknowledge seqr and the broad institute)?
• How does CPU thread scaling work? Does it automatically scale up/down the threads? I did see a setting that can change Gunicorn worker threads here though.
• When users upload their own vcfs, will these vcfs be automatically annotated?

[bpblanken]

Hi @YasirKusay,

Responses below:

1. Yes, you should be able to use a browser and navigate to localhost provided you port-forward from kubernetes. After installation on kubernetes, you might see:

> kubectl get services
NAME                       TYPE        CLUSTER-IP      EXTERNAL-IP   PORT(S)                                        AGE
hail-search                ClusterIP   10.51.242.126   <none>        5000/TCP                                       68d
kubernetes                 ClusterIP   10.51.240.1     <none>        443/TCP                                        5y141d
seqr                       NodePort    10.51.248.72    <none>        8000:30950/TCP                                 68d
seqr-redis-headless        ClusterIP   None            <none>        6379/TCP                                       68d
seqr-redis-master          ClusterIP   10.51.244.59    <none>        6379/TCP

You should then run kubectl port-forward service/seqr 8000:8000 to map the kubernetes service port 8000 your own device's port 8000. You should be able to visit localhost:8000 in your browser and see the seqr user interface.

2. Yes, only http is provided by default. However, we encourage implementation of ssl with an Ingress resource. You may enable and configure an Ingress with the setting here.

3. The gunicorn worker threads argument sets the number of seqr application_ processes that can run. It is a static number and is recommended to be set at (2 * #CPU)+1.

4. We unfortunately do not support users uploading their own VCFs at the moment; the administrator needs to manage the VCF upload prior to initiating the loading pipeline that annotates the VCF.

[YasirKusay]

Hi @bpblanken,

Thank you for the detailed responses, just some clarification. For point 1, do you have a guide for how to configure http (asides from just enabling ingress in this file). For point 3, could you clarify what you mean by the number of seqr application_processes? Is it the annotation scripts that can run simultaneously? In addition, are there other settings where we can adjust the number of threads? For point 4, I feel like it goes against the first new feature here. Doesn't the first new feature imply that users can upload data, which will be automatically annotated, etc.

One more question, is it possible to restrict who can access seqr by limiting access to certain IP addresses.

[bpblanken]

Hi @YasirKusay,

1. Unfortunately we don't have a guide for the Ingress. Internally we're using the Ingress-Nginx controller, so I'd recommend looking at the documentation here. There are also notes here relating to ssl.

2. The threads relates to the seqr Django application only (web server requests and responses, standard CRUD flows, auth etc.) and does not impact the loading pipeline. A reasonable of setting of 9 should scale the application server to several dozen concurrent users.

3. The specific piece that isn't supported is the network upload from the client (or other remote location) to the server that's running seqr. Once the VCF is uploaded to the loading datasets directory, annotation is trigger-able from the UI and completely automated.

4. Yes, this is definitely possible. There are likely many ways to accomplish this though, one being at the Kubernetes Network Policy layer, depending on the infrastructure preferences of your organization.

[YasirKusay]

Thank you @bpblanken, this has been incredibly helpful!