[Feature] support Cassandra(Docker) as backend (compose with HugeGraph)

[aroundabout]

Feature Description (功能描述)

subtask of [Feature] Support build docker image from Dockerfile (Task Summary) #840

Key Problem:

1. In order to init_store after the cassandra is up, the && ./bin/init-store.sh in Dockfile should be moved to start-hugegraph.sh.
2. We should wait for the storage (cassandra) being up.
3. In order to change the 'hugegraph.properties', we should pass environment variable according to docker-compose.yaml. And use an env_var_parser to parse them.

Main process:

1. Move the init_store.sh from dockerfile to start-hugegraph.sh

2. Use the env var to change the hugegraph.properties. Provide a shell which can process environment variables whose keys start with "hugegraph." and updates a property file accordingly. In this way, we can easily pass the env var to the properties, and if we want to support other backend, we need no additional modification in the code. Shell is like:

   while IFS=' ' read -r envvar_key envvar_val; do
       if [[ "${envvar_key}" =~ hugegraph\. ]] && [[ ! -z ${envvar_val} ]]; then
           envvar_key=${envvar_key#"hugegraph."}
           if grep -q -E "^\s*${envvar_key}\s*=\.*" ${GRAPH_PROP}; then
               sed -ri "s#^(\s*${envvar_key}\s*=).*#\\1${envvar_val}#" ${GRAPH_PROP}
           else
               echo "${envvar_key}=${envvar_val}" >> ${GRAPH_PROP}
           fi
       else
           continue
       fi
   done < <(env | sort -r | awk -F= '{ st = index($0, "="); print $1 " " substr($0, st+1) }')
   

   the part of the demo compose file:

   services:
   graph:
     build: /home/dandelion/incubator-hugegraph/
     # image: hugegraph/hugegraph:latest
     container_name: ca-graph
     ports:
       - 18080:8080
     environment:
       hugegraph.backend: cassandra
       hugegraph.serializer: cassandra
       hugegraph.cassandra.host: ca-cassandra
       hugegraph.cassandra.port: 9042
     networks:
       - ca-network
     depends_on:
       - cassandra
   

3. Wait for storage backend initialization, then init_store

   demo shell like:

   if ! [ -z "${HUGE_STORAGE_TIMEOUT_S:-}" ]; then
       F="$(mktemp --suffix .groovy)"
       echo "graph = HugeFactory.open('${GRAPH_PROP}')" > $F
       timeout "${HUGE_STORAGE_TIMEOUT_S}s" bash -c \
       "until bin/gremlin-console.sh -e $F > /dev/null 2>&1; do echo \"waiting for storage...\"; sleep 5; done"
       rm -f "$F"
   fi
   "${BIN}/init-store.sh"
   

   use 'org.apache.tinkerpop.gremlin.groovy.jsr223.ScriptExecutor' in gremlin-console.sh to execute the graph = HugeFactory.open('${GRAPH_PROP}'). If success or overtime, exec init-store.sh

4. demo docker-compose is like:

version: "3"

services:
  graph:
    build: /home/dandelion/incubator-hugegraph/
    # image: hugegraph/hugegraph:latest
    container_name: ca-graph
    ports:
      - 18080:8080
    environment:
      hugegraph.backend: cassandra
      hugegraph.serializer: cassandra
      hugegraph.cassandra.host: ca-cassandra
      hugegraph.cassandra.port: 9042
    networks:
      - ca-network
    depends_on:
      - cassandra

  cassandra:
    image: cassandra:3.11
    container_name: ca-cassandra
    ports:
      - 7000:7000
      - 9042:9042
    security_opt:
      - seccomp:unconfined
    networks:
      - ca-network
    healthcheck:
      test: ["CMD", "cqlsh", "--execute", "describe keyspaces;"]
      interval: 10s
      timeout: 30s
      retries: 5

networks:
  ca-network:

volumes:
 hugegraph-data:

Risk:

1. I am not sure if I can move the init_store.sh from dockerfile to start-hugegraph.sh

Others

We can provide a link in dockerhub doc, which link to the docker-compose.yaml in the main repo. I think it is a better way for user to use docker-compose. File structure is like:

./hugegraph-dist
├── docker
│       └── example
│               └── docker-compose-cassandra.yml
│               └── other-template-docker-compose.yml