CLI
Karafka has a simple CLI built in. It provides the following commands:
| Command | Description |
|---|---|
| help | Describe available commands |
| console | Start the Karafka irb console similar to the Rails console (short-cut alias: "c") |
| info | Print configuration details and other options of your application |
| install | Installs all required things for Karafka application in current directory |
| server | Start the Karafka server (short-cut aliases: "s", "consumer") |
| swarm | Start the Karafka server in the swarm mode (multiple forked processes) |
| topics | Allows for topics management (create, delete, repartition, reset, migrate, align, plan) |
| topics health | Check Kafka topics for replication and durability issues (Pro) |
All the commands are executed the same way:
bundle exec karafka [COMMAND]
To list all available commands with a short description of each, run:
bundle exec karafka help
Karafka server¶
Limiting consumer groups used per process¶
Karafka supports having multiple consumer groups within a single application. You can run multiple Karafka instances, specifying consumer groups that should be running per each process using the --include-consumer-groups server flag as follows:
bundle exec karafka server --include-consumer-groups group_name1,group_name3
If you specify none, by default, all will run.
You can also exclude certain consumer groups by using the --exclude-consumer-groups flag:
bundle exec karafka server --exclude-consumer-groups group_name2,group_name3
Limiting subscription groups used per process¶
Karafka supports having multiple subscription groups within a single application. You can run multiple Karafka instances, specifying subscription groups that should be running per each process using the --include-subscription-groups server flag as follows:
bundle exec karafka server --include-subscription-groups group_name1,group_name3
If you specify none, by default, all will run.
You can also exclude certain subscription groups by using the --exclude-subscription-groups flag:
bundle exec karafka server --exclude-subscription-groups group_name2,group_name3
Handling Multiplexed Subscription Groups in CLI Commands
Multiplexed subscription group replicas share the same name; multiplexing does not add any postfix to it. Use the subscription group's own name, the same one you would use without multiplexing, when including or excluding it via CLI flags.
Limiting topics used per process¶
Karafka supports having multiple topics within a single application. You can run multiple Karafka instances, specifying topics that should be running per each process using the --include-topics server flag as follows:
bundle exec karafka server --include-topics topic_name1,topic_name3
If you specify none, by default, all will run.
You can also exclude certain topics by using the --exclude-topics flag:
bundle exec karafka server --exclude-topics topic_name2,topic_name5
Using Wildcard Patterns¶
The --include-* and --exclude-* server flags for consumer groups, subscription groups and topics accept wildcard patterns in addition to literal names. Karafka treats a value as a pattern when it contains *, ?, [ or ], and matches it like a file name glob. A value without these characters must match the name exactly.
# Run all consumer groups except those whose names start with app-a-
bundle exec karafka server --exclude-consumer-groups 'app-a-*'
# Mix patterns and literal names
bundle exec karafka server --include-topics 'orders_*',payments
Karafka does not check wildcard patterns against your routing, so a pattern that matches nothing does not cause an unknown name error. If no topics remain to subscribe to, the server still refuses to start. When a name matches both an inclusion and an exclusion, inclusion wins, the same as with literal names.
With Routing Patterns, topic filters match a pattern by its name, the same as literal names do, so a wildcard can include or exclude whole patterns.
Karafka Topics Health¶
topics health Is a Pro Command
The topics health command is part of Karafka Pro.
The topics health command analyzes your Kafka topics for replication and durability issues. It inspects each topic's replication factor and min.insync.replicas setting to detect potential risks.
bundle exec karafka topics health
The command checks for the following conditions:
- No Redundancy - Topics with a replication factor of 1, meaning no replicas exist. Any single broker failure will cause data loss.
- Zero Fault Tolerance - Topics where the replication factor is less than or equal to
min.insync.replicas, meaning no broker can fail without causing the topic to become unavailable for writes when producers useacks=all. - Low Durability - Topics where
min.insync.replicasis set to 1, meaning acknowledged writes usingacks=allonly require a single broker, increasing the risk of data loss if that broker fails before replication completes.
Each finding is printed as it is found (in cluster order, not grouped by severity), shown in red for critical issues or yellow for warnings. After all topics are checked, if any issues were found, a generic recommendations summary is printed, covering increasing the replication factor, raising min.insync.replicas, and keeping the replication factor above min.insync.replicas for fault tolerance. If no issues were found, a single "All topics are healthy" confirmation is printed instead.
Karafka Info¶
The info command prints the Karafka, Ruby and karafka-rdkafka versions, the consumer, share and subscription group counts, the workers count, the client id, the boot file, the environment and the license details:
bundle exec karafka info
Add the --extended flag to also print the routing tree, the global application config and the effective Kafka config:
bundle exec karafka info --extended
The extended output adds three sections:
- Routing - each consumer group and share group with its state (active or inactive) and its active features. For each subscription group: its name, position, state,
kafka[group.id],max_messages,max_wait_timeand, when multiplexing is used, itsmin,maxandbootvalues. For each topic: its consumer,max_messages,max_wait_time,initial_offset,consumer_persistence, pause settings, its ownkafkasettings (when set) and active features. - Config - the global settings:
client_id,group_id,concurrency,max_messages,max_wait_time,initial_offset,shutdown_timeout,consumer_persistence, pause settings,strict_topics_namespacingandstrict_declarative_topics. - Kafka Config - the global
kafkasettings.
Sensitive Kafka Values Are Redacted
Karafka prints [REDACTED] instead of the value of each kafka setting whose name contains password, passphrase, secret, token, sasl.oauthbearer.client, private.key.pem, ssl.key.pem or ssl_key (case-insensitive). This applies to the Kafka Config section and to the per-topic kafka settings. Karafka does not redact other values, so ssl.certificate.pem and ssl.ca.pem are printed.
Karafka Swarm¶
Swarm has its own section. You can read about it here.
Declarative Topics¶
Declarative Topics managament via the CLI has its own section. You can read about that here.
Routing Patterns¶
Routing Patterns managament via the CLI has its own section. You can read about that here.
See Also¶
- Getting Started - Initial setup and basic CLI usage
- Deployment - Using CLI commands in production environments
- Env Variables - Environment variables affecting CLI behavior
Last modified: 2026-11-11 22:26:52