The situation with update lag keeps deteritoriating (it's 2 hours behind now and is not improving) and it looks like we've reached the bottleneck for capacity. The servers with no load seem to be keeping up fine, but the loaded ones keep falling behind.
- Lower query timeout
- Add more servers to serve the load
- Reduce throttling thresholds
It looks like the update load has increased significantly recently, and we have to keep up somehow.
Any other ideas are welcome. We have a long-term plan to look into update performance within Blazegraph, but it will probably take significant time to develop something working, and in the meantime we have servers crumbling under the load.