leader-elect will create a temporal lock in etcd, setting the hostname
of the running leader. The leader will periodically reset the ttl while
other nodes will periodically check whether the key is still there and
race to become leader when that's not the case.
The logic is inspired by kubernetes podmaster.
Configuration is done using the command line arguments and environment
variables. When using the provided systemd units, it will load
configuraton from /etc/leader-elect/global and /etc/leader-elect/KEY
though.
leader-elect [OPTIONS] SERVICE
| Option | Description | Default value |
|---|---|---|
| --etcd-servers=[] | etcd cluster | http://localhost:2379 |
| --ttl=30 | ttl for the lock in seconds | 30 |
| --sleep=5 | time between checking the lock | 5 |
| --whoami=HOSTNAME | value to use for the lock | the hostname |
| --unit=systemd-unit | systemd unit to start/monitor | SERVICE.service |
| --key | key to use for the lock | /leader-elect/SERVICE |
Options need not be supplied on the command line. They can be provided
using environment variables where the option name is uppercased, -
replaced by _ and MASTER_ELECT_ or MASTER_ELECT_SERVICE_ is
prepended. E.g. for leader-elect my-service it will look into
MASTER_ELECT_TTL and MASTER_ELECT_MY_SERVICE_TTL to determine the ttl.
Commandline arguments take preference over the specialised environment variables and those take preference over the general environment variables.
leader-elect is distributed under the terms of the GPL-3. See COPYING. If that might be a problem for you, contact me.