Awesome

btsync.docker

A 32MB docker image sync service built for persist data on clusters

+------+    +----------+
|  NFS |----| Your App |
+------+    +----------+

Imagine a NFS server, but made for cluster and containers.

Reliability

Ready for production! Until now, its the only BTSync docker imaged that is harassed with a test suite!

Motivation

CLUSTER PERSISTENT STORAGE

At findhit.com, we have to share data between our cluster nodes. After testing out some solutions such as: cloud provider's storage ones, some well known cluster-targeted file systems and among other less valuable solutions to our architecture (such as NAS sharing), we've got always the same opinion. None of them could be fast, scalable and easy as plug-n-play.

As so, we looked for p2p syncing solutions. BTSync was our first choice, but it lacks some important features such as cluster-based configuration storages (etcd for example).

I took a time to plan how I would structure configurations on a cluster, but it always ends on configurations per node, the only difference with current BTSync approach would be configuration sharing, and thats not useful.

So, instead of creating an image that relied on confd (which is great btw), I've decided to work on btsync cli before going to sleep.

How does it works?

A cli will allow us to place a btsync service per cluster node (making use out of the ephemeral disk they mostly have) and control that service via docker exec.

Service structure example

NOTE: Examples presented above will rely on a CoreOS environment.

• Submitting a btsync global service (with fleet) will allow us to share one path (my choice was /mnt/resource, which is an ephemeral storage on CoreOS on MS Azure)
• Services could easily ensure or remove a folder by running:
  • docker exec btsync ctl add gitlab on ExecStartPre
  • docker exec btsync ctl del gitlab on ExecStopPost
• Storage could be mounted by sharing /data volume of btsync container, but I really advise you to don't do so. /data could have other containers data and that fact creates a big SECURITY WARNING ON MY HEAD!!! Instead we could use a path getter such as: docker run -v $(docker exec btsync ctl path gitlab):/home/gitlab/data gitlab:latest

Usage

There are several ways to configure this service. They were created to fit on your services structure.

Note: Although they don't mention how to bind data folder into a disk, you could achieve by mounting it as a volume: -v /path/to/some/host/disk:/data

Launching btsync container

As a NFS-redundant Stack Service (RECOMMENDED)

docker run \
    --name btsync-nfs \
    bootstrap SECRET_HERE some-third-party-image

# Now you just have to make sure your image is able to mount it
# In case the image you pretend doesn't do that, you could work around it by
# replacing entrypoint and command. (Advanced users)
docker run \
    --name some-third-party-image \
    -e CONTAINER_DATA_PATH=/container/data/path \
    --entrypoint /bin/sh \
    some-third-party-image: latest \
    -s -c "mount -t nfs4 btsync-nfs:/some-third-party-image /container/data/path && /path/to/original/entrypoint and/or command;"

Shares are exposed on NFS root system, meaning that if you create a namespace called yolo (with ctl add yolo), you nfs mount command should look like:

mount -t nfs4 btsync.link.or.ip:/yolo /path/to/mount/on

As a Global Service way

docker run \
    --name btsync \
    cusspvz/btsync:latest

Now each service should add his folder data before mounting it:

docker exec btsync ctl add --secret="SECRET_HERE" my-app-name
docker run --name="my-app-name" -v /path/to/some/host/disk/my-app-name:/container/data/path my-app-name

# Removing it from global service
docker exec btsync ctl del my-app-name

As a sidekick volumes-from

docker run \
    --name my-app-name-data \
    cusspvz/btsync:latest \
    bootstrap SECRET_HERE my-app-name

docker run \
    --volumes-from my-app-name-data \
    -e DATA_DIR=/data/my-app-name \
    my-app-name

Accessing BTSync WebGUI

BTSync WebGUI is available by default at port 8888.

You could set up an internal load-balancer just to see whats happening under the hood. But please, avoid to give public access to it, otherwise MR Robot will nuke all your data, including you chinese servers backup with a Raspberry Pi...

Default credentials: User: admin Pass: admin

Commands

Running commands on container

docker exec btsync ctl [commands]

add [--secret=""] namespace

Adds a folder for a determined namespace

docker exec btsync ctl add --secret="e3hryu35qegqery4w5y164u5u" "gitlab"

NOTE: Secret provided on example isn't a valid one, it was generated by a random pseudo-geek key typing. I'm on a coffee, now people here think I write fast.

del namespace

Deletes folder related with that namespace

docker exec btsync ctl del "gitlab"

path [--scope=(host|container)] namespace

Returns host's or container's path to created namespace. By default, scope is host.

docker exec btsync ctl path "gitlab" # /mnt/resources/sync/gitlab

has namespace

Exits with 0 (has) or 1 (hasn't) indicating if namespace exists [or not]

if docker exec btsync ctl has "gitlab"; then
    echo "yeah, we already have that namespace"
fi

list

Prints a spaced-separated list of namespaces

docker exec btsync ctl add gitlab
docker exec btsync ctl add star-trek
docker exec btsync ctl list # gitlab star-trek

Environment Variables

HOST_DATA_PATH

Defaults to: /mnt/resources

DATA_PATH

Defaults to: /data

CONFIG_PATH

Defaults to: /data/btsync.conf

CONFIG_INTERVAL_CHECK

Defaults to: 10

PID_PATH

Defaults to: /var/run/btsync.pid

DEBUG

Defaults to: "btsync:ctl"

UNAME

Defaults to: btysnc

UID

Defaults to: 1000

GID

Defaults to: 1000

Development

build and run a container

make run

build and run a container with bash command

make run-bash

just building a container

make build

build image, build test image, run multiple containers and harass them with test suite

make test