Improvements to Performance and Response Time
FC-6394 - Significantly improved overall Myst response time by up to 90x by reducing the number of unnecessary sql queries
- Improved all page load times - Most pages load within seconds as opposed to 10+ seconds or minutes
- Minimised hourly CPU spike - Hourly executions of Release Pipelines reduced from minutes to a second thus reducing CPU time
- Expect cases where large datasets are loaded (ie. thousands of stream model properties) to take several seconds however, no longer take minutes to load
This improvement deserves a special thanks to Rob de Haan (Rabobank) who was very pragmatic and cooperative in working through the issues. Rob gave valuable data which contributed to the successful outcome. Here is an example of how the Platform Model loading has improved illustrated by the new load time (blue) and the previous (red).
The issue relates to how relational entities were retrieved from the database for every workspace resource. The improved result can be seen in the graph showing response times between 6.6.0-rc8 and previous versions. Note this particular graph is based off very large data.
Bug Fixes
- FC-6386 - Fixed a regression bug where release pipeline could not be edited and stages could not be added (existing pipelines)
Known Issues
We are looking to resolve these issues in a future release candidate.
Platform Model Unexpected Automatic Update
Platform Models are automatically and unexpectedly updated by Myst under the following conditions. Note the last point is what triggers the unexpected update.
- Release Pipeline stream's behaviour is set to Update a blueprint
- Release Pipeline stage deploy scheme is set to Automatic deployments
- (Trigger) A new revision is created against the platform model (ie. a user save and commits a platform model change).
Unable to Access Terminated Platform Models
Platform Models that are in a TERMINATED state will not open and return an error. This issue was identified with a particular customer. We are yet to see this occur on other datasets. A root cause analysis is underway.