cqllock

README

cqllock

Command-line tool to create and delete distributed locks using Cassandra and lightweight transactions. These tools are useful for managing distributed jobs on multiple hosts.

Usage

Set up a config file at ~/.cqllockrc or /etc/cqllock.yaml in the following YAML format:

seeds:
- cassandra-seed1.domain.com
- optional-cassandra-seed2.domain.com
- etc
keyspace: keyspace_containing_lock_table
table: lock_table_name
username: optional-username-if-using-auth
password: optional-password-if-using-auth
certpath: /path/to/optional/client/cert/if/using/SSL.pem
keypath: /path/to/optional/client/key/if/using/SSL.key
timeout: 10  # This is the Cassandra connectiont timeout in seconds
retries: 3   # This is the number of times to retry failed Cassandra queries

Possible scenarios

Ensure a process only runs once ever, regardless of how many times or on what host it's attempted:

cqllock ProcessLock && ./myProcess.sh

The first call to cqllock that successfully acquires the lock will hold it forever. Further calls to cqllock will return non-zero, so ./myProcess.sh won't run again. This command can be run on multiple hosts.

Multiple hosts run a job, but only allow one at a time:

cqllock -t 1d -r 5m jobLock && ./myJob.sh ; cqllock -u jobLock

The first host to acquire the lock will hold it until ./myJob.sh finishes. Other hosts will attempt to acquire the lock every five minutes and will wait up to one day trying to reaquire the lock.

There are likely other useful scenarios.