The data cache is not balanced

[liyangbing]

use juiceFS sync pull 2.2T data from external OSS to the JuiceFS data directory from a machine
the data cache of the pull machine accounts for almost 2T, and the other 15 machines are hardly used
is that how it works? Or is there something wrong with the Settings
Performance should be limited by disk reads and writes on a single machine

What you expected to happen:
Data is evenly distributed to each data cache node

Environment:
Juicefs: 1.0.0 - beta 2
Operation : 4.15.0-171- Generic #180-Ubuntu
Metadata: Redis primary and secondary
Number of data caches machine: 16

[davies]

The data cache is built separately, not shared with each other, the size of cached data is determined by workload on them. So this is expected.

[dannyfe]

Is there a plan to build a shared cache?

[davies]

Yes, there is a PR: #1364