mirror of
https://github.com/kvcache-ai/AgentENV.git
synced 2026-11-05 01:17:36 +00:00
[GH-ISSUE #264] rootfs space explosion #74
Labels
No labels
bug
documentation
enhancement
enhancement
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
kvcache-ai/AgentENV#74
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @cheney-xiaobo on GitHub (Sep 8, 2026).
Original GitHub issue: https://github.com/kvcache-ai/AgentENV/issues/264
Affected component
Snapshot / template builder
Problem statement
When continuously overwriting a file inside the sandbox, the
upper.datafile underfirecracker‑work/agentenv‑fc‑xxfor overlaybd keeps growing and can reach up to 10 GB. If multiple pause/resume operations are superimposed in this scenario, the overall rootfs footprint can become extremely large, then snapshot may can not finish(because snapshot will not compact rootfs overbd.commit)Use case
in firecracker write 1G file 10 times,then pause sandbox/resume sandbox, continue many time
Current behavior and workarounds
before compact work the rootfs overlaybd.commit may reach 10GB
Desired behavior
maybe there should be compact before snapshot or there is some other good way to solution this issue
Proposed approach
No response
Compatibility and operational impact
API/config changes:
add manual compact method or do compact before snapshot with many layer
Snapshot or storage format changes:
New host/runtime requirements:
Upgrade and rollback considerations:
Alternatives considered
or change the HybridLogStructured mode
Non-goals
No response
Contribution
I can implement it
Pre-submission checklist
@yingdi-shan commented on GitHub (Sep 8, 2026):
We have a built-in compaction mechanism after 32 pause/resumes. We are working on optimizing it in issue #183.