Problem
Container creation currently isn’t recoverable. If the process crashes or is interrupted mid-way, the only option is to delete everything and start from scratch. This is painful for large containers.
Desired behavior
If container creation is interrupted, rerunning the command should resume from the existing partially-created container state instead of rebuilding everything.
Proposed approach
Store a small, machine-detectable header/metadata record for each container “block” (chunk) that includes:
- Block coordinate
- Block size
- Optional integrity info (e.g., checksum / hash)
On startup, the creation process should:
- Detect, that the block index is invalid or missing and then:
- Scan existing blocks in the target container
- Validate blocks (presence + expected size, optionally checksum)
- Reuse valid blocks
- Regenerate missing or invalid blocks only
- Continue until the container is complete
- write the correct block index
Problem
Container creation currently isn’t recoverable. If the process crashes or is interrupted mid-way, the only option is to delete everything and start from scratch. This is painful for large containers.
Desired behavior
If container creation is interrupted, rerunning the command should resume from the existing partially-created container state instead of rebuilding everything.
Proposed approach
Store a small, machine-detectable header/metadata record for each container “block” (chunk) that includes:
On startup, the creation process should: