Skip to content

Latest commit

 

History

History
58 lines (39 loc) · 3.7 KB

README.md

File metadata and controls

58 lines (39 loc) · 3.7 KB
Raw

Embulk as a Micro-service

This project aims to facilitate deploying embulk as a micro-service through SSH tunneling

What is does

  1. Connect to your database 1
  2. Do a job like converting from Db1 to Db2 (as specified in the configuration_example.yml file)
  3. Connect to your database 2 and write the Embulk output

Every connection is done using SSH tunneling.

Example, with Mongo (database 1) and Postgres (database 2) :
example

Can it be on hosted on PAAS ?

Yes, you can host it on heroku for instance, or on your own server.

How can I install it ?

Pre requisite : you need Docker installed on your machine

Then, you have to :

  • put your ssh key (the private part) in the .ssh folder as keyexample or default_env_SSHKEY according to your environment_variables you will define next step => this key will allow this machine to connect to the remote database so you need also to make sure the remote machines will allow the connection with a public key
  • customize the environment variables in the environment_variables.txt file according to the different IP of your servers etc...
  • modify configuration_example.yml according to your needs (see embulk website for more details)
  • run docker build --build-arg CONFIGURATION_FILE=configuration_example.yml --build-arg DIFF_FILE=diff.yml --tag embulk_container . to launch the build process of your docker image
  • run docker run --env-file=environment_variables.txt -it embulk_container bash only later if you want to start the process again. If you change environment_variables.txt of your configuration_example.yml you will need to run the other one in order to build again the docker image

Note :

  • For better use, I suggest renaming configuration_example.yml to configuration.yml and since it is gitignored you can leave it in the repo. Another example can be found named configuration_example_2.yml
  • for incremental update, we need to keep "diff.yml" (see embulk doc) from one run to another. In order to do so, we set up a Docker Volume to keep it persistent. This is donc adding -v $PWD:/work to the docker run command. So here is the command:
    docker run --env-file=environment_variables.txt -v $PWD:/work -it embulk_container bash
  • If, for some unkwnown reason, you cannot merge the first time, try to insert instead, and manually specify the primary key on your output database
  • you may encounter some database error Sort operation used more than the maximum XXXXXX bytes of RAM in case of incremental_field while you haven't indexed your database on this field
  • using the java:8 docker image was triggering an out of RAM problem. we switched to this image FROM fabric8/java-jboss-openjdk8-jdk:1.4.0 in order to have the ability to limit Java Ram usage docker run -m 600m -e JAVA_OPTIONS='-Xmx300m' [...]. This issue was inherent to Java, unable to use cgroup memory limits : whatever the container Ram limit was, Java container was using all the machine ressource, causing big errors.
  • when running in production, don't forget to remove -it because no TTY will ba available if you trigger it from a CRON job for instance

TroubleShooting

  • If you still get prompt password, you have an issue with your SSH auth, It can be that your key has too wide permission. try
chmod 600  .ssh/keyexample

Examples:

From Mongo to Postgres

see this example

From Mongo to Postgres, with transformation

see this example

From Postgres to BigQuery

see this example