The framework forms the basis of a production kdb+ system by implementing some core functionality and utilities on top of kdb+, allowing developers to concentrate on the application business logic. It incorporates as many best practices as possible, with particular focus on performance, process management, diagnostic information, maintainability and extensibility. Wherever possible, we have tried to avoid re-inventing the wheel and instead have used contributed code from code.kx.com (either directly or modified). This framework will be suitable for those looking to create a new kdb+ system from scratch or those looking to add additional functionality to their existing kdb+ systems.
Have a skim through our brochure for a bit more information. The easiest way to get a production capture started is to download and install one of the Starter Packs, or read through our Github-Pages site. We also have a Google Group for questions/discussions.
Quick Start
To launch a process wrapped in the framework, you need to set the environment variables and give the process a type and name. The type and name can be explicitly passed on the command line. setenv.sh is an example of how to set the environment variables on a unix type system. For a windows system, see http://www.computerhope.com/issues/ch000549.htm. kdb+ expects all paths to be / (forward-slash) separated so all paths on all OSs should be forward-slash separated.
To avoid standard out/err being redirected, used the -debug flag
./setenv.sh /- Assuming unix type OS
q torq.q -proctype test -procname mytest -debug
To load a file, use -load
q torq.q -load mytest.q -proctype test -procname mytest -debug
This will launch the a process running within the framework with all the default values. For the rest, read the document!
Updating the Documentation with Mkdocs
To make changes to the documentation website you must simply use this command while in the branch you have made the changes on:
mkdocs gh-deploy
You will be prompted to enter a username and password, after this the site should have been updated. You can test the site locally if you want using mkdocs. First use the command:
mkdocs build
Then:
mkdocs serve -a YourIp:Port
Head to the address it gives you to check if your changes have worked. More information about using mkdocs can be found here
Release Notes
- 4.2.0, Dec 2021
- Added functionality to Data Access API
- Added Data Striping
- Bug fixes
- Improved testing
- 4.1.1, Jun 2021
- Data Access API bug fixes
- Improved attribute routing surrounding overlapping processes
- 4.1.0, Apr 2021
- Data Access API Update Blog post and Documentation
- 4.0.1, Mar 2021
- Bug fixes
- Documentation improvements
- Test improvements
- 4.0.0, Dec 2020
- Updated documemtation
- Segmented Tickerplant
- Test Framework
- 3.8.0, Oct 2020
- Updated documentation
- Install script added
- Inclusive terminology
- DQS,DQC and DQE upgrades
- PCAP functionality additions
- 3.7.0, May 2020
- Added Data Quality System (blog post and documentation)
- Fix k4unit testing so that it works with kdb+ 4.0.
- 3.6.0, Apr 2020:
- Added DataDog integration for monitoring: https://www.aquaq.co.uk/q/monitoring-a-torq-system-with-datadog/
- Improved k4unit integration
- Tested with kdb+ 4.0
- Fix for reval calls via gateaway
- Fix for timeout at EOD in RDB when -T specified
- Various minor improvements and bug fixes
- 3.5.0, Aug 2019:
- .torq stop and .torq summary functions updated
- .torq restart function added
- Grafana updated
- Monitor checks script added
- Email disabled if trying to run on Win64
- Opencon modified to append user:password to tcps connections
- Bug fixes
- 3.4.1, Jan 2019:
- Gateway Query .gw.syncexec bug identified and fixed
- 3.4.0, Dec 2018:
- Update of the torq.sh script to include debug and print start lines.
- Integration of Monit for monitoring TorQ.
- Integration of Grafana adaptor, allowing visualization of TorQ process data.
- Added service layer for client specific customisation across all aplications.
- Bug Fixes including:
- Ability to specify parentproctype for processes when starting with torq.q
- Function to check .proc.getattributes for non-TorQ processes.
- Exit when setenv not sourced
- qcon now allowed in debug mode
- Hanging when null result returned from aysnc gateway request
- 3.3.0, May 2018:
- Upgraded to run with kdb+ v3.6. Note that it remains backwardly compatible with prior versions.
- Implemented concurrent synchronous requests, as part of 3.6 upgrade
- Added Response Formatter for gateway
- Added Low Power Mode - aimed at use with Kx On Demand Licence
- Added Initialisation Function invoked after a script is loaded and connections are established.
- Changed the start script, to start or stop processes separately, in a batch or all at once. All the process configuration is now at one place in $KDBCONFIG/process.csv.
- Bug fix for erroneous brackets that were causing permissions config to be loaded in incorrect order
- 3.2.0, February 2018:
- Added initial version of analytics library, a set of functions to aid data manipulation
- Added optional dependency checker for applications to check if they have minimal TorQ and kdb+ dependencies
- Added kdb compression to filealerter. Use case is for compression of tickerplant log files, so log files can be replayed directly rather than having separate uncompress step
- Added -onelog command line option to write stdout/stderr to same log file
- Added file pattern option to data loading utility
- Bug Fixes
- 3.1.0, May 2017:
- added kafka which is a real time messaging system with persistent storage in message logs
- added datareplay.q with functionality for generating tickerplant function calls from historical data which can be executed by subscriber functions
- added subscribercutoff.q with functionality that can be used to cut off slow subscribers from processes
- added new write down method for tickerlogreplay so match the writedown methods in the WDB
- 3.0.0, January 2017:
- Added a permissioning system, allowing granular control of access to users & groups
- Added LDAP support, allowing a user to authenticate against an LDAP server
- Improved documentation now available at http://aquaqanalytics.github.io/TorQ/
- 2.7.0, November 2016:
- Tickerplant incorporated. Tickerplant has faster recovery for real time subscribers, and better timezone handling
- Filealerter uses a splayed table to store the table of already processed data. If it finds a table in current format (flat) it will change it to a splay
- Gateway bug fixes
- Small improvements
- 2.6.2, September 2016:
- .z.pd and peach logic added to wdb.q
- Bug fixes
- 2.6.0, August 2016:
- Added broadcast publishing
- Added domain sockets and tcps as ipc connection mechanisms
- Added fallback from domain sockets to tcp functionality
- New gateway (which allows attributes to be sent rather than specifying the processes to hit)
- Heartbeat subscriptions are easier
- Supports snappy compression
- Tested with kdb+ 3.4
- Bug fixes
- 2.5.0, April 2016:
- Added u.q - publish/subscribe code from KDB+ Tick
- Improved memusage.q to do sampling
- Updated dataloader.q
- Make sure all config can be read from KDBAPPCONFIG
- 2.4.0, February 2016:
- Added k4unit to allow tests to be automatically run via the -test parameter
- Extended application config handling so all default config can be found in $KDBAPPCONFIG
- Bug Fixes
- 2.3.0, December 2015:
- Added optional write down method to wdb process, which allows data to be written to custom partition schemes during the day. At the end of day before being moved to the hdb, data is merged instead of sorted, which will allow the data to be accessed sooner. The optional method may present a significant saving in time for datasets with a low cardinality (small distinct number of elements), i.e. FX market data
- Added write access control to message handlers (using reval), which restricts the ability of querying clients to modify data in place
- Added functionality to return approximate memory size of kdb+ objects
- Bug fixes
- 2.2.1, November 2015:
- Bug fix - fixed endofdaysort to not throw type error when par.txt is available
- Bug fix - fixed gateway so results are not dropped when a client loses connection and another is querying multiple servers
- 2.2.0, October 2015:
- Added application configuration management using $KDBAPPCONFIG environmnet variable
- Added pid, host and port to heartbeat table
- Changed gateway to be non intrusive
- Bug fixes
- 2.1.0, July 2015:
- Added a chained tickerplant process
- Updated housekeeping.csv to take in an extra column agemin which represents whether to use minutes or days in find function
- Updated email libraries
- 2.0.1, May 2015:
- Added RDB process which extends r.q from kdb+ tick.
- Added WDB to write down data periodically throughout the day. Extends w.q.
- RDB and WDB allow seamless end-of-day event (no data outage, no tickerplant back pressure)
- Added Reporting Process to run reports periodically and process the results
- Added environment variable resolution to process.csv to allow greater portability. If a process is started without a port specified it will look it up from process.csv based on the proctype and procname.
- Added -localtime flag to allow process to run in localtime rather than GMT (log message, timer calls etc.). The change is backwardly compatible - without -localtime flag the process will print logs etc. in GMT but can also have a different .z.P
- Added Subscription code to manage multiple subscriptions to different data sources
- Added email library which uses libcurl. Used to send emails from TorQ processes
- Added standard monitoring checks to the database code
- Added data loader script. Utility functions to load a directory of data into a database in chunks, sort and part at the end
- Added tickerplant log recovery utilities to recover as many messages as possible from a log file rather than just stopping at the first bad message
- Added compression process to run and compress a given database
- Modified compression code to handle par.txt databases
- Modified compression code and housekeeping process to run with kdb+ 2.*
- Modified std out/err logging and usage logging to include process name and process type (the logmsg table had changed along with some of the functions in the .lg namespace so you might need to check in case you have overridden any of them)
- Removed launchtick scripts and some default configuration: to create a test system, install a starter pack
- 1.2.0, Sep 2014:
- Tested on kdb+ 3.2
- Added connections to external (non TorQ) processes using nonprocess.csv
- Modified file alerter with optional switch to move or not move a file if any function fails to process the file
- Discovery service(s) host:port(s) can be passed on the command line (.servers.DISCOVERY) to a process (this should enable complete bypassing of process.csv if required)
- Add custom hook (.servers.connectcustom) which is invoked whenever a new connection is made (allows, for example, subscription to a new process)
- Add optional application detail file ($KDBCONFIG/application.txt) to allow customisation of the start up banner (application version etc.)
- If required env. variables (KDBCODE, KDBCONFIG, KDBLOG) are not set they will default to $QHOME/code, $QHOME/config, $QHOME/logs respectively (previously the process failed and exited)
- 1.1.0, Apr 2014:
- Added compression utilities, HTML5 utilities, housekeeping process, file alerter process, kdb+tick quick start
- 1.0.0, Feb 2014:
- Initial public release of TorQ