Skip to content

Latest commit

 

History

History
236 lines (183 loc) · 9.57 KB

README.md

File metadata and controls

236 lines (183 loc) · 9.57 KB

Docver docker-compose deployment

Docver consists of six components, each of which is defined in base/<component_dir>. Both base and template directories should not be edited and you are expected to configure your deployment in a separate directory. To configure this we provide init.sh to easily create one or multiple deployment instances and configurations

For more information on the topology of the services that make the solution please refer to documentation in helm package provided here.

Prerequisites

  • docker compose >= 2.22.0
  • docker >= 20.10.5

Docker compose

To install docker compose, follow the instructions here.

If you don't want to "pollute" your docker environment we recommend using a standalone installation of docker-compose. It is a single binary that you can put in your PATH and use it as a regular command as explained in the provided link.

Your licence key

As mentioned here. You can acquire these on developer.microblink.com. If you are using Document verification self-hosted you will need to acquire the licence key from the licences section (under Document verification self-hosted).

Quickstart - Deploying your instance

Clone this repository and position yourself in the root of the repo: git clone [email protected]:microblink/doc-ver-ops.git && cd doc-ver-ops

Initialise your environment and running the deployment

To initialize your environment you should use init.sh provided in the docker-compose directory.

You can do that by running the following commands

cd docker-compose
bash init.sh <deployment_name> <licence_key> <application_id>

Update the values in <> with your own values.

  • <deployment_name> - the name of your deployment, we recommend simply using your company name
  • <licence_key> - the licence key that you will use to authenticate. This key is acquired on developer.microblink.com under licences - document verification self-hosted.
  • <application_id> - application identifier. This value is acquired on developer.microblink.com under licences - document verification

This will create a new directory in the docker-compose directory with the <deployment_name> you provided. This directory will contain the configuration files for your deployment runnable by a single command docker-compose up -d.

To prevent any confusion, if for example I'm working for company microblink, my licence key is someExampleString and application id is myapp I would run the following command:

bash init.sh microblink someExampleString myapp

Afterwards, I would continue to deploy my instance by running the following:

cd microblink
docker-compose up -d

Inspecting the logs

To inspect the logs of a specific service, run docker-compose logs <service_name>.

Available services

  • doc-ver-api
  • doc-ver-runner
  • visual-anomaly
  • retrieval-intermediary
  • embedding-store-server
  • seeder
  • pgvector

Startup

To successfully startup, our solution needs to initialize its database. Depending on the database size this can be a lengthy process. The database initialization is done by the seeder service. The seeder service will run once and initialize the database. After the database is initialized, the seeder service will exit.

To track the progress simply use docker-compose logs seeder, there you will see the logs on percentage of the database that is initialized.

Where is the Document verification endpoint available?

Once initialized, the Document verification endpoint will be available on the machine you started this docker-compose deployment on, on port 8080.

If you wish to change the port, you can do so by updating the configuration of the doc-ver-api service in the in the config/doc-ver-api/.env file, by updating the value of DOC_VER_API_PORT variable.

Available databases

The database used can be configured at conf/embedding-store-server/seeder.yaml by changing the gc-seed-store-prefix field. By default we use full-db as the prefix.

Right now, we only have the full-db and croatia-db databases available. Using croatia-db is helpful to confirm that the deployment is initializing and working as expected as it is by an order of magnitude smaller than full-db. Do not use croatia-db for production environments as it is likely not representative of the real-world use case.

Resource requirements

For each service you will be able to configure how much CPU and RAM it can use. Alongside docker-compose.yaml there is a conf directory. Inside the conf directory there is a directory for each service. Inside each service directory there is a .env file where you can configure the resource requirements for the service.

We set the default values for each service in the conf directory. Please reach out to us if you need to change the default values as we need to make sure that the services are not starved of resources. Given your expected load we can help you determine the right values for your deployment.

Configuration details

We do not expect the need to change the default configuration for each service. However, if you need to change the parts of our deployment that are not exposed via env vars (that you can configure in .env files) you can do so by extending them in //docker-compose.yaml. Since creds and conf are loaded outside the /base directory, //.env file needs to be added. Inside there should be an env variable with the full path to the / directory on the target server.

Server prerequisites and configuration options for each component will be described in the following sections.

Prerequisites

  • doc-ver-api licence key

Components

doc-ver-api

doc-ver-api docker-compose.yaml will start one longrunning doc-ver-api server service

conf

To configure doc-ver-api, an .env file should be added to /<deployment_dir>/conf/doc-ver-api, with the following values:

ALLOWED_ORIGINS=<string_value>
LIMITS_CPUS=<int_value>
LIMITS_MEM=<string_value>
RESERVATIONS_CPUS=<int_value>
RESERVATIONS_MEM=<string_value>
creds

Not needed for doc-ver-api

doc-ver-runner

doc-ver-runner docker-compose.yaml will start longrunning doc-ver-runner server service. By default one process will be started, you can increase number by setting REPLICAS parameter.

conf

To configure doc-ver-runner, an .env file should be added to /<deployment_dir>/conf/doc-ver-runner, with the following values:

APPLICATION_ID=<string_value>
LIMITS_CPUS=<int_value>
LIMITS_MEM=<string_value>
RESERVATIONS_CPUS=<int_value>
RESERVATIONS_MEM=<string_value>
REPLICAS=<int_value>

creds

To be able to work, doc-ver-runner needs a license key that you can provide in /<deployment_dir>/creds/doc-ver-runner:

LICENSE_KEY=<string_value>

visual-anomaly

visual-anomaly docker-compose.yaml will start one longrunning visual-anomaly server service

conf

To configure visual-anomaly, an .env file should be added to /<deployment_dir>/conf/visual-anomaly, with the following values:

ALLOW_CORS=false
LIMITS_CPUS=2
LIMITS_MEM=2Gb
RESERVATIONS_CPUS=1
RESERVATIONS_MEM=1Gb

creds

Not needed for visual-anomaly

retrieval-intermediary

retrieval-intermediary docker-compose.yaml will start one longrunning retrieval-intermediary server service

conf

To configure retrieval-intermediary, an .env file should be added to /<deployment_dir>/conf/retrieval-intermediary, with the following values:

COLLECTION_NAME=mdv
PARALLEL_QUERIES=100

creds

Not needed for retrieval-intermediary

embedding-store

embedding-store docker-compose.yaml will start three services, pgvector database, seeder and embedding-store-server. seeder inserts predefined data to server database and exits when done. embedding-store-server and pgvector are longrunning services.

seeder

conf

To configure seeder for s3, seeder.yaml should be added to /<deployment_dir>/conf/embedding-store, with the following values:

grpc-recv-size: <int_val>
grpc-send-size: <int_val>

col-create-workers: <int_val>
col-insert-workers: <int_val>
col-insert-batch: <int_val>

s3-seed-store: <bool_val>
s3-seed-store-endpoint: <string_val>
s3-seed-store-secure: <bool_val>
s3-seed-store-bucket: <string_val>
s3-seed-store-prefix: <string_val>

To configure seeder for gcs, seeder.yaml shold be added to /<deployment_dir>/conf/embedding-store, with the following values:

grpc-recv-size: <int_val>
grpc-send-size: <int_val>

col-create-workers: <int_val>
col-insert-workers: <int_val>
col-insert-batch: <int_val>

gc-seed-store: <bool_val>
gc-seed-store-bucket: <string_val>
gc-seed-store-prefix: <string_val>

server

conf

To configure server and db, .env file should be added to /<deployment_dir>/creds/embedding-store, with following values:

GRPC_RECV_SIZE=<int_value>
GRPC_SEND_SIZE=<int_value>
GRPC_TIMEOUT=<string_value>

PG_VECTOR_CMX_CONN=<int_value>
PG_VECTOR_LIMITS_MEM=<string_value>
PG_VECTOR_RESERVATIONS_MEM=<string_value>
creds

Not needed for embedding-store server

bundle

To configure bundle services, see base/bundle/README.md (Services section)