Tracking issue for forest-cli archive

[lemmih]

Issue summary

In order to generate and maintain the archival data described here, we need better tooling for manipulating archives. This tooling is independent of the daemon (i.e., does not require a running node) and should operate directly on CAR files.

• forest-cli archive export --epoch=X #3159
• forest-cli archive export --epoch=X --diff=Y #3204
• forest-cli archive snapshot --epoch=X --spine
• forest-cli archive snapshot --epoch=X --diff=Y --spine
• forest-cli archive union
• forest-cli archive info #3166

Other information and links

[ruseinov]

Let's flesh out forest-cli archive snapshot --epoch=X.

I'll outline the process the way I understand it

1. We have a car file
2. We index a car file using a car-backed Blockstore
3. We then export a snapshot up to a certain tipset at --epoch=X
4. How big of a readable window the snapshot has to represent. e.g. if we export at --epoch=X, should --epoch=x-900 still be verifiable (with all the lookups and such)
5. If exported epoch is not present in the storage - bail.

[lemmih]

Spot on.

For point 4, let's make it a flag with 900 epochs as the minimum and 2000 epochs as the default (to align with Lotus).

[lemmih]

@ruseinov I created a separate issue for the task: #3159

[ruseinov]

@ruseinov I created a separate issue for the task: #3159

Perfect!