feat: sealing: StartEpochSealingBuffer triggers packing on timer

[ZenGround0]

Proposed Changes

Improvements to snap deals integration beyond the existing mvp PR

• use the config's StartEpochSealingBuffer param as a way to enforce that sectors are packed for sealing / updating no matter how many deals they have if the nearest deal start date is close enough to the present. This solves an existing problem predating snap deals (FYI @jennijuju)
• Config option for disabling automatic creation of new sectors for handling incoming deals. Forces user to trigger inclusion by marking sectors for snap upgrade.
  - [ ] Recycle assigned deals in updating sectors that have been aborted so that these deals can be reassigned to other sectors

[jennijuju]

Nice - I assume it is intentional that this will apply to regular deal sectors too?

the remaining two todos would be really nice to have - I think we should aim this for lotus v1.15.0 (Jan 18). (I feel like its a bit tight to get this implemented and tested in v1.14.0, which code freezes in 2 days. its a super nice enhancement to the currently sealing pipeline - so okay to wait for v1.15.0)

[jennijuju]

Please do **not ** merge this PR without a pr to https://lotus.filecoin.io/docs/storage-providers/config/

[ZenGround0]

It turns out that recycling deals from a failed sector is a more involved change than I had thought and therefore is going to have to wait beyond v 1.15.0. I am going to file an issue which I'll link to from back here describing the problem a bit more.

--edit--
This is well covered in filecoin-project/boost#1345 where I've identified some problems

[magik6k]

Looks good, just 2 small nits

[magik6k]

I don't think we need this copied variable, can just use sealTime directly

[BlocksOnAChain]

@ZenGround0 can we merge it for 1.15.0, since I think we got positive feedback from Magik?
CC: @jennijuju

[codecov]

Codecov Report

Merging #7905 (485568d) into master (899ae8a) will decrease coverage by 0.04%.
The diff coverage is 50.00%.

@@            Coverage Diff             @@
##           master    filecoin-project/lotus#7905      +/-   ##
==========================================
- Coverage   39.12%   39.08%   -0.05%     
==========================================
  Files         660      660              
  Lines       71459    71474      +15     
==========================================
- Hits        27961    27938      -23     
- Misses      38681    38710      +29     
- Partials     4817     4826       +9     

Impacted Files	Coverage Δ
extern/storage-sealing/input.go	53.89% <38.46%> (-0.66%)	⬇️
node/config/def.go	97.48% <100.00%> (+0.01%)	⬆️
node/modules/storageminer.go	63.56% <100.00%> (+0.05%)	⬆️
chain/actors/builtin/miner/diff.go	48.52% <0.00%> (-10.30%)	⬇️
markets/retrievaladapter/client_blockstore.go	62.50% <0.00%> (-6.25%)	⬇️
miner/warmup.go	70.83% <0.00%> (-4.17%)	⬇️
chain/types/tipset_key.go	85.96% <0.00%> (-3.51%)	⬇️
extern/sector-storage/sched_worker.go	77.32% <0.00%> (-2.33%)	⬇️
storage/miner.go	69.31% <0.00%> (-2.28%)	⬇️
miner/miner.go	56.06% <0.00%> (-1.97%)	⬇️
... and 17 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 899ae8a...485568d. Read the comment docs.

[jennijuju]

oh btw! bad pr title!! reminder to follow the one in pr template next time

[ZenGround0]

@jennijuju You'll need to help me read the template better because I tried my best to follow it.

PR type: :

The title is feat: sealing: * description of change being made* right?

PR type: fix, feat, INTERFACE BREAKING CHANGE, CONSENSUS BREAKING, build, chore, ci, docs,perf, refactor, revert, style, test

feat

area: api, chain, state, vm, data transfer, market, mempool, message, block production, multisig, networking, paychan, proving, sealing, wallet, deps

sealing

Are the underscores what you are talking about? Double checking they're not being used in other PRs.

[jennijuju]

(just to close the loop from slack) what you said is all ✅

when I left the comment, the title was " StartEpochSealingBuffer triggers packing on timer " and I update it!