if use write back then upload large file very slow

[lygn128]

What happened: when mount --write-back then dd write large file speed change from 500MB/s to 100+ MB/s

What you expected to happen:

How to reproduce it (as minimally and precisely as possible):
juicefs mount --writeback(or not)
dd if=/dev/zero of=./$(hostname).40G bs=4096000 count=1000

Anything else we need to know?

Environment:

• JuiceFS version :juicefs version 1.4.0-dev+2025-07-03.c7dadcb7
• Cloud provider or hardware configuration running JuiceFS:
• OS (e.g cat /etc/os-release): ubuntu 22.04
• Kernel (e.g. uname -a):5.15.0-156-generic
• Object storage (cloud provider and region, or self maintained): aws s3
• Metadata engine info (version, cloud provider managed or self maintained): redis 6
• Network connectivity (JuiceFS to metadata engine, JuiceFS to object storage):
• Others:

[jiefenghuang]

With write-back enabled, blocks are first written to local disk and then uploaded. For large files this is not recommended.
You can upload directly to S3 instead.
If you’d like to dig into the issue further, you can use the JuiceFS performance-analysis tools (stats etc.) and common utilities to pinpoint the bottleneck.

[jiefenghuang]

Maybe we can add more control to writeback, so that it can filters based on the size of the write.