setup arax-backup.rtx.ai server (a backup server for the ARAX ARA andd KG2 API)

[saramsey]

This idea came out of the mini-hackathon today.

[saramsey]

I have provisioned an r5a.2xlarge instance (Ubuntu 18.04 AMI) named arax-backup.rtx.ai in us-west-2 with a root filesystem in a 1023 GiB EBS volume (AWS instance ID i-0f3589ffb64a0fb20). For the instance network security policy, I enabled security groups for inbound HTTPS (TCP port 443), HTTP (TCP port 80), and SSH (TCP port 22).

[saramsey]

I assigned elastic IP address 44.236.84.186 to this instance.

[saramsey]

I set up a DNS A record arax-backup.rtx.ai pointing to IP 44.236.84.186.

[saramsey]

I temporarily updated the DNS A record arax.rtx.ai to point to IP 44.236.84.186.

[saramsey]

I updated apt-get inside the instance and installed certbot, python-certbox-nginx, and emacs inside the instance (this installs nginx by dependency).

[saramsey]

I set up @finnagin's RSA public key inside /home/ubuntu/.ssh/authorized_keys inside the instance. I also generated a public key /root/.ssh/id_rsa.pub using ssh-keygen and installed that public key in /home/sramsey/.ssh/authorized_keys on arax.ncats.io.

[saramsey]

Inside the instance, I made a new directory /mnt/data and, using sftp -C, I copied the following files into it from arax.ncats.io:/translator/data/orangeboard/databases/KG2.7.3:

node_synonymizer_v1.0_KG2.7.3.sqlite 
meta_kg_v1.0_KG2.7.3c.json 
kg2c_v1.0_KG2.7.3.sqlite 
fda_approved_drugs_v1.0_KG2.7.3c.pickle 
curie_to_pmids_v1.0_KG2.7.3.sqlite 
config_local.json 
autocomplete_v1.0_KG2.7.3.sqlite 
GRAPH_v1.0_refreshed_KG2.7.3.sqlite DTD_probability_database_v1.0_refreshed_KG2.7.3.db 
COHDdatabase_v1.0_KG2.7.3.db

[saramsey]

From the arax.ncats.io instance, I copied the config file /etc/nginx/sites-enabled/default to /home/ubuntu/default on arax-backup.rtx.ai. I edited the file changing arax.ncats.io to arax.rtx.ai throughout, and I changed /etc/nginx/dhparams.pem to /etc/ssl/certs/dhparam.pem. I installed that updated config file in /etc/nginx/sites-enabled/default on arax-backup.rtx.ai (the original version can be found in /etc/nginx/sites-available/default).

[saramsey]

I generated a new file dhparam.pem by running (on arax-backup.rtx.ai, as root):

cd /etc/ssl/certs && openssl dhparam -out dhparam.pem 4096

[saramsey]

I started nginx using

sudo service nginx start

and tested that TLS works for https://arax-backup.rtx.ai and https://arax.rtx.ai. TLS works fine, but of course, per the config file, it is TLS-terminated and redirects to port 8080 which is not currently working (yet) because we haven't set up the Docker container for ARAX, so we get a 502 HTTP status code (Bad Gateway Error).

[saramsey]

Having verified that https://arax.rtx.ai works (connects to the backup server), I then updated the DNS A record for arax.rtx.ai to point back to the IP address for arax.ncats.io, since under normal circumstances we want arax.rtx.ai to point to the arax.ncats.io server.

[saramsey]

Tagging @amykglen @finnagin so they are aware of the current status of the issue.

[saramsey]

Copying over the synonymizer_artifacts subdirectory now...

[saramsey]

OK, I believe I have copied over all the relevant files from /translator/data/orangeboard/databases/KG2.7.3 on arax.ncats.io to /mnt/data/orangeboard/databases/KG2.7.3 on arax-backup.rtx.ai.

[saramsey]

So @finnagin you have the green-light to install ARAX on arax-backup.rtx.ai.

[finnagin]

Update: I tried spinning up the merged dockerfile but ran into a few issues that may take a while to figure out how to fix. Since we are low on time I'm going to just spin up a container from an image of the container hosted on arax.ncats.io. Though, because this will take a little while and I have to shut down arax.ncats.io to do it I will wait until this evening to make the image.

[finnagin]

I believe I now have arax-backup.rtx.ai up and running.

[finnagin]

Though currently I cannot test the UI as when I try to connect via the url arax-backup.rtx.ai I get a 403 error.

[saramsey]

Relay is over. arax-backup.rtx.ai is working, but now the instance is stopped to save AWS costs (we can just start the instance if we ever need to run the backup ARAX).