exoscale/zlocker

Zookeeper based isolated command execution

zlocker: zookeeper based isolated command execution

⚠️ Usage of this tool is deprecated and being replaced by consul lock. Please check the Consul Lock Documentation for details on how to use it.

zlocker is a small tool meant to ease sequential execution of a command across a large number of hosts.

Configuration

zlocker is configured through command line arguments:

-z
    Comma-separated list of zookeeper servers to contact
-l
    Name of lock node in zookeeper
-w
    Optional sleep period, defaults to none
-t
    Zookeeper session timeout

The rest of the command line will be fed to the shell if a lock is successfully acquired.

Building

If you wish to inspect zlocker and build it by yourself, you may do so by cloning this repository and peforming the following steps :

go build

Example usage

zlocker -z zk01,zk02,zk03 -l /zlock-service-restart service my-daemon restart

zlocker's Issues

zlocker eats first command

Seems zlocker eats the first command instruction:

$ zlocker -z `cat /etc/zlocker.cluster.conf` -l /zlock-test -w 5 -- sh -c "echo start; sleep 30; echo end"
zlocker: Connected to 
zlocker: Authenticated: id=216172784299502539, timeout=40000
zlocker: Re-submitting `0` credentials after reconnect
end
command finished, sleeping

'echo start' wasn't executed. If I add it a second time:

$ zlocker -z `cat /etc/zlocker.cluster.conf` -l /zlock-test -w 5 -- sh -c "echo start; echo start; sleep 30; echo end"
zlocker: Connected to 
zlocker: Authenticated: id=144660276477914487, timeout=40000
zlocker: Re-submitting `0` credentials after reconnect
start
end
command finished, sleeping

cancel execution of a command if it fails on one host

It could be very useful to have a way to stop the execution of a command on hosts after it fails on one host.

• don't release lock if command fails
• add an execution timeout so if a host can't get a lock after thus duration, execution is cancelled