You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Lightning creates thousands of empty regions. Lightning hit the store limit very often and it sometimes lead to skewed regions.
We should skip the store limit if scatter on empty regions are inexpensive.
The text was updated successfully, but these errors were encountered:
If we don't scatter the regions or the scatter operation is pending/canceled, the import job may have hotspots on TiKV. Most of the import flow might happen on the three TiKVs if we only use the 3 replicas setting, which affects its progress and the data might be skewed on TiKVs.
Lightning with parallel import can cause split operations on non empty regions. I want to understand the implication of that.
bufferflies
changed the title
skip store limit for emptry regions created by lightning
skip store limit for empty regions created by lightning
Dec 30, 2024
Enhancement Task
Lightning creates thousands of empty regions. Lightning hit the store limit very often and it sometimes lead to skewed regions.
We should skip the store limit if scatter on empty regions are inexpensive.
The text was updated successfully, but these errors were encountered: