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
When a compaction task contains a lots of SSTs as input, it incurs bursting object store reads due to 1) meta block prefetch and 2) opening streaming read for SSTs. It can result in compaction task failure due to some resource limits (which I still cannot figure out #7002): risingwave_storage::hummock::compactor: Compaction task 2198903 failed with error: ObjectStore failed with IO error timeout: error trying to connect: HTTP connect timeout occurred after 3.1s."
To Reproduce
Let compactor executes a large compaction task, which contains 1000+ small SSTs.
zwang28
changed the title
Avoid large concurrent object store reads caused by too many SSTs in compaction task input
Avoid bursting object store reads caused by too many SSTs in compaction task input
Mar 15, 2023
Describe the bug
When a compaction task contains a lots of SSTs as input, it incurs bursting object store reads due to 1) meta block prefetch and 2) opening streaming read for SSTs. It can result in compaction task failure due to some resource limits (which I still cannot figure out #7002):
risingwave_storage::hummock::compactor: Compaction task 2198903 failed with error: ObjectStore failed with IO error timeout: error trying to connect: HTTP connect timeout occurred after 3.1s."
To Reproduce
Let compactor executes a large compaction task, which contains 1000+ small SSTs.
Expected behavior
No response
Additional context
https://risingwave-labs.slack.com/archives/C03DLHWHEN4/p1678857503219889
The text was updated successfully, but these errors were encountered: