Cleanup Calibration datasets to (temporarily) reduce chain bandwidth pressure before GA

[BigLep]

Done Criteria

~50% of the ~11k datasets on calibration as of 2026-02-03 are deleted.

Why Important

1. Reduce the strain on Calibration as discussed in Audit impact of increasing DEFAULT_MAX_PROVING_PERIOD on Calibnet #409
2. Helps validate that users can self-service delete their datasets and pieces

Notes

1. Some quick analysis should be done to see the top creator of datasets to ensure that cleaning up dealbot and loadtests will get us the headroom we are looking for.
2. We don't just want to terminate datasets as termination results in 30 more days of proving by default. We want to also delete pieces.
3. The suggested set of actions would be: - List all datasets
   • For each dataset
     • List all pieces
     • Delete each piece
     • Terminate the dataset
4. filecoin-pin already has rm --piece functionality per Add rm --piece $pieceCid command filecoin-project/filecoin-pin#8. Terminating datasets is in progress with Add data-set $dataSetId --terminate filecoin-project/filecoin-pin#189. Those should be the key things someone needs to script the above.
   • Bonus if someone wants to create a rm --dataset to do all of the above in one command: feat(rm): add --all option to remove all pieces from DataSet filecoin-project/filecoin-pin#320

[BigLep]

Note: I came up with the ~50% number in the done criteria. I assume that's a good target, but if someone has a more informed number, please share.

[beck-8]

1960+84+187=2231
I am deleting these number of data sets(and pieces).

[BigLep]

@beck-8 : thanks.

Can you also share the commands you're running for reference to do the deletes, termination, and verification that it worked...

@silent-cipher : can you look to see how much datasets you can free up from the legacy dealbot?

[beck-8]

filecoin-project/filecoin-pin#320
I submitted a PR. However, I actually used James’ script and asked AI to modify it for me.
https://github.com/beck-8/synapse-sdk/blob/feat/dataset-tool/utils/dataset-tools.js

[silent-cipher]

can you look to see how much datasets you can free up from the legacy dealbot?

I don’t think I’ll be able to free up much from the legacy dealbot. Datasets were terminated earlier without deleting pieces, and now they’re past end epoch, so the contract condition here blocks cleanup.
Feels like there should be auto-cleanup path in Curio for terminated datasets - happy to be corrected if I’m missing a case.