Doccano on Azure

Deployment

Follow the doccano-deploy.sh script to deploy your resource. We need to have an Azure Ressource Group and a running Posgres server in it. It required the following env variables in your shell (fill the profile.sh file from the profile.sh.template).

• you choose the admin username, password and email for doccano
  • ADMIN_USERNAME
  • ADMIN_PASSWORD
  • ADMIN_EMAIL
• you find the information about the existing resources
  • SUBSCRIPTION_ID
  • POSTGRES_SERVER_NAME (az postgres server list --resource-group $RG_NAME | grep fullyQualifiedDomainName, you also can parse it with jq)
  • CREATION_POSTGRES_ADMIN_USERNAME
  • POSTGRES_SERVER_PASSWORD
  • DATABASE_NAME
• when you have deployed your doccano App Service, report the URL in the profil.sh

Activate

source profile.sh

or

make activate

Launch a new panai retro game

make new-panai-retro

WIP Get answers

make get-retro-panai-answers

Unit testing

\n

make tests

make coverage

Scale out

To handle the workload before a massive session :

On Azure portal/App Service resource : Scale-up

To reduce the bill amount between two sessions :

On Azure portal/App Service resource : Scale-up

Whitelist the app in the postgres db

Since the postgres database has firewall rules, we need to whitelist our App Service (doccano) resource. We need to do it again after a scale-up / scale-down because the machines are switched.

Get the outbound addresses of the App Service resource

And whitlist it in the postgreSQL database.

Either manually, or...

Get the outbound ips of the App Service resource

export OUTBOUND_IP_ADDRESSES=$(az webapp show --resource-group $RG_NAME --name $WEB_APP_NAME --query outboundIpAddresses --output tsv)

Whitelist all of them in th PostgreSQL resource

count=0
for i in $(echo $OUTBOUND_IP_ADDRESSES | tr "," "\n")
do
  echo "$i\n"
  az postgres server firewall-rule create --start-ip-address $i --end-ip-address $i --name doccano${count} --resource-group $RG_NAME --server-name qmpostgresql
  let "count+=1" 
  echo "$count\n"
  echo "doccano${count}"
done

TODO : if we terraform this infrastructure, this can be simplified.

Stress test

In order to test if your application is well scaled and will handle the workload. You can do a stress test. The following test create 100 of simulataneous connections and use the admin account to annotate and un-annotate an image a 100 of times.

$ python stress_test.py

And follow the metric you want (response time for example) on Azure portal App Service resource/Monitoring.