This container might be used to bring up Zookeeper standalone cluster. Container has complicated configuration generation using confd and serf. Serf is used to solve zookeeper's lack of dynamic configuration (as of 3.4.8 version), this means that ensemble members should be known before start up which causes chicken-egg problem for a docker environment.
- Serf cluster is brought up, containers wait when all members of ensemble appear to scale up (go online).
- Once the required number of members appeared and they ip configuration determined configs generation takes place.
- Zookeeper service is brought up on line.
| Variable | Description | Default |
|---|---|---|
| SERF_JOINTO | Specifies the host address when starting the second and the following zookeeper instances. | |
| ENSEMBLE_SCALE_TIMEOUT | Ensemble scale up timeout. Specifies deadline for all ensemble members to go online. | 300 sec |
| ZK_ENSEMBLE_SIZE | Required number of Zookeeper instances which form quorum. | 3 |
| ZK_ADDRNUM | Specifies ip address number which Zookeeper will be using, by default the first interface. | 1 |
| ZK_TICK_TIME | Sets tickTime. | 2000 ms |
| ZK_INIT_LIMIT | Sets initLimit. | 10 ticks |
| ZK_SYNC_LIMIT | Sets syncLimit. | 5 ticks |
| ZK_MAX_CLIENT_CXNS | Sets maxClientCnxns. | 60 |
# Start containers interactively (from different consoles).
docker run -d --name zookeeper-a actionml/zookeeper
# wait for start (follow logs and ^C)
docker logs -f zookeeper-a
# Grab ip address of an instance to join to (Serf join).
zk_ip=$(docker inspect -f '{{ .NetworkSettings.IPAddress }}' zookeeper-a)
# Start as many instances as required, by default you have to have 3, so +2.
docker run -d -e SERF_JOINTO=$zk_ip actionml/zookeeper
docker run -d -e SERF_JOINTO=$zk_ip actionml/zookeeper
As soon as the ensemble converges the quorum will be printed out bellow ==> Zookeeper quorum:, it can be pasted somewhere into another configuration (for example HBase).
- Denis Baryshev (dennybaa@gmail.com)