capillaries

Capillaries

Capillaries is a data processing framework that:

• addresses scalability issues and manages intermediate data storage, enabling users to focus on data transforms and quality control;
• bridges the gap between distributed, scalable data processing/integration solutions and the necessity to produce enriched, customer-ready, production-quality, human-curated data within SLA time limits.

This is a GitHub readme. More details and a blog at https://capillaries.io.

Why Capillaries?

Getting started

On Mac, WSL or Linux, run in bash shell (make sure to copy data files before starting containers):

git clone https://github.com/capillariesio/capillaries.git
cd capillaries
./copy_demo_data.sh
docker compose -p "test_capillaries_containers" up

Navigate to http://localhost:8080 to see Capillaries UI, wait until it shows the Keyspaces screen with no errors. It may take a while - all docker containers must start and Cassandra must be fully initialized. Now Capillaries is ready to process sample demo input data according to the sample demo scripts (all copied by copy_demo_data.sh above).

Start a new Capillaries data processing run by clicking "New run" and providing the following parameters (no tabs or spaces allowed in textboxes):

Alternatively, you can start a new run using Capillaries toolbelt by executing the following command from the Docker host machine, it should have the same effect as starting a run from the UI:

docker exec -it capillaries_webapi /usr/local/bin/capitoolbelt start_run -script_file=/tmp/capi_cfg/portfolio_quicktest/script_quick.json -params_file=/tmp/capi_cfg/portfolio_quicktest/script_params_quick_fs_one.json -keyspace=portfolio_quicktest -start_nodes=1_read_accounts,1_read_txns,1_read_period_holdings

Watch the progress in Capillaries UI. A new keyspace portfolio_quicktest will appear in the keyspace list. Click on it and watch the run complete - nodes 7_file_account_period_sector_perf and 7_file_account_year_perf should produce result files:

cat /tmp/capi_out/portfolio_quicktest/account_period_sector_perf.csv
cat /tmp/capi_out/portfolio_quicktest/account_year_perf.csv

Monitoring your test runs

Besides Capillaries UI at http://localhost:8080, you may want to check out the stats provided by other tools.

Log messages generated by:

• Capillaries Daemon
• Capillaries WebAPI
• Capillaries UI
• Message queue
• Cassandra with Prometheus jmx-exporter
• Prometheus are collected by fluentd and saved in /tmp/capi_log.

To see Cassandra cluster status, run this command (reset JVM_OPTS so jmx-exporter doesn't try to attach to the nodetool JMV process):

docker exec -e JVM_OPTS= capillaries_cassandra1 nodetool status

Cassandra read/write statistics and some Daemon/Webapi metrics collected by Prometheus available at:

http://localhost:9090/query?g0.expr=sum%28irate%28cassandra_clientrequest_localrequests_count%7Bclientrequest%3D%22Write%22%7D%5B1m%5D%29%29&g0.show_tree=0&g0.tab=graph&g0.range_input=15m&g0.res_type=auto&g0.res_density=medium&g0.display_mode=lines&g0.show_exemplars=1&g1.expr=sum%28irate%28cassandra_clientrequest_localrequests_count%7Bclientrequest%3D%22Read%22%7D%5B1m%5D%29%29&g1.show_tree=0&g1.tab=graph&g1.range_input=15m&g1.res_type=auto&g1.res_density=medium&g1.display_mode=lines&g1.show_exemplars=0&g2.expr=irate%28capi_script_def_cache_hit_count%5B1m%5D%29&g2.show_tree=0&g2.tab=graph&g2.range_input=15m&g2.res_type=auto&g2.res_density=medium&g2.display_mode=lines&g2.show_exemplars=0&g3.expr=irate%28capi_script_def_cache_miss_count%5B1m%5D%29&g3.show_tree=0&g3.tab=graph&g3.range_input=15m&g3.res_type=auto&g3.res_density=medium&g3.display_mode=lines&g3.show_exemplars=0

Further steps

Blog at capillaries.io

For more details about this particular demo, see Capillaries blog: Use Capillaries to calculate ARK portfolio performance. To learn how this demo runs on a bigger dataset with 14 million transactions, see Capillaries: ARK portfolio performance calculation at scale.

Further introduction

For more details about getting started, see Getting started.

Deploy Capillaries at scale

Container-based deployments

Capillaries binaries are intended to be container-friendly. Check out the docker-compose.yml, it may be a good starting point for creating your full-scale container-based deployment.

VM-based deployment

See Terraform script that creates Capillaries deployment in AWS.

Capillaries in depth

What it is and what it is not (use case discussion and diagrams)

Getting started (run a quick Docker-based demo without compiling a single line of code)

Testing

Toolbelt, Daemon, and Webapi configuration

Script configuration

Capillaries UI

Capillaries API

Glossary

Q & A

Capillaries blog

MIT License

(C) 2022-2025 KH (kleines.hertz[at]protonmail.com)