Cluster administration

riffle-ctl queries coordinator tables and applies administrative changes to individual or piped server instances.

Command groups

riffle-ctl query ...
riffle-ctl instance { set | add | remove | kill }
riffle-ctl bench { disk { append | read | profile } | hdfs append }
riffle-ctl inspect file ...
riffle-ctl config init ...

Query applications

./riffle-ctl query \
  --coordinator http://coordinator.example.com:21001 \
  --sql "select * from active_apps"

./riffle-ctl query \
  --coordinator http://coordinator.example.com:21001 \
  --sql "select * from historical_apps"

Update one server

./riffle-ctl instance set \
  --instance 192.168.1.10:19998 \
  --status unhealthy

Replace the server’s custom tags:

./riffle-ctl instance set \
  --instance 192.168.1.10:19998 \
  --tags sata,rack-a

Built-in protocol tags such as ss_v4, ss_v5, and GRPC are always retained.

Pipeline mode

When --instance is omitted, instance commands read JSON Lines from standard input. Every row must include ip and http_port.

./riffle-ctl query \
  --coordinator http://coordinator.example.com:21001 \
  --sql "select * from instances where tags like '%sata%'" \
  --json \
  | ./riffle-ctl instance set --status unhealthy

Pipeline mode makes the selection explicit and reviewable: run the query by itself first, inspect the resulting instances, and then pipe the same query into the update.

Confirm the coordinator endpoint, SQL filter, and selected instances before running instance set, add, remove, or kill against a production cluster.