I was testing the brain split brain issue of TDengine these days with different configurations which prevents me from continuing to deploy cluster. In order to reduce the possibility of encountering bottlenecks, the unusable scenarios of TDengine are necessary to declare in advance. So I want to know what situation will cause the TDengine cluster core dump or unavailable except split brain or master selection failure? How can TDengine cluster effectively ensure high availability?
I think TDengine is designed for storing/processing time-series data, for example in IoT industry data collected by sensor or data generated through monitoring devices over time. The underlying data structure for storing data on Disk uses LSM tree so its more suitable for situations with more data writing than reading since new data will be just append to the last. Basically TDengine is more suitable for storing time-series data with respect to optimizing the performance, but not recommended for usage in OLAP and other situations.