Closed VitaliStupin closed 6 years ago
This same behavior is also seen during performance testing made by Palveluväylä testing team. Results are similar as shown above (there are occasional 'peaks' in response times). This bug is related how the time-stamping is done. Current time-stamping implementation is not correctly supporting the parallel execution of code but is working more or less in a serialized way.
Partial index related to time-stamping (see https://github.com/vrk-kpa/xroad-joint-development/issues/120 ) is slightly improving the functionality but not is the completely solution. Implementation of time-stamping should be refactored/redesigned to support parallel operations.
Fixed in 6.17.0
Affected components: - Affected documentation: - Estimated delivery: - External reference: -
Problem During load tests it was noticed that periodically message exchange stops as if timestamping would lock messagelog table and prevent new messages to be sent.
To illustrate the problem. Chart with timestamps (timestamping made once a minute on each of two security servers):![withtimestamps](https://cloud.githubusercontent.com/assets/15190150/25806171/09544580-340b-11e7-8cf9-dea74a05050f.png)
Chart without timestamps:![withouttimestamps](https://cloud.githubusercontent.com/assets/15190150/25806176/0d7a846c-340b-11e7-8aa8-a34a0f0ddfaf.png)
Acceptance criteria