blockstore: atomize slot clearing, relax parent slot meta check

[AshwinSekar]

clear_unconfirmed_slot can leave blockstore in an irrecoverable state
if it panics in the middle. write batch this function, so that any
errors can be recovered after restart.

[codecov]

Codecov Report

Attention: Patch coverage is 93.33333% with 5 lines in your changes are missing coverage. Please review.

Project coverage is 81.8%. Comparing base (6ee3bb9) to head (bc0f888).
Report is 8 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff            @@
##           master   #35124    +/-   ##
========================================
  Coverage    81.7%    81.8%            
========================================
  Files         834      834            
  Lines      224299   224828   +529     
========================================
+ Hits       183361   183923   +562     
+ Misses      40938    40905    -33     

[AshwinSekar]

reworked as per #35124 (comment)
the writes now happen atomically

[mergify]

Backports to the stable branch are to be avoided unless absolutely necessary for fixing bugs, security issues, and perf regressions. Changes intended for backport should be structured such that a minimum effective diff can be committed separately from any refactoring, plumbing, cleanup, etc that are not strictly necessary to achieve the goal. Any of the latter should go only into master and ride the normal stabilization schedule.

[mergify]

Backports to the beta branch are to be avoided unless absolutely necessary for fixing bugs, security issues, and perf regressions. Changes intended for backport should be structured such that a minimum effective diff can be committed separately from any refactoring, plumbing, cleanup, etc that are not strictly necessary to achieve the goal. Any of the latter should go only into master and ride the normal stabilization schedule. Exceptions include CI/metrics changes, CLI improvements and documentation updates on a case by case basis.

[steviez]

We don't have to do this sanity checking if we fetch the SlotMeta in this function ourselves instead of relying on the caller

Technically, this would add some overhead since we'd be fetching the SlotMeta twice, once here and once in Blockstore::clear_unconfirmed_slot(). I think that's ok, but if we're conscious about that, I think we could remove the SlotMeta check in Blockstore::clear_unconfirmed_slot() anyways. The rocksdb range deletes are cheap, and this function should only be getting called if ReplayStage knows it has something that it needs to delete.

That being said, fetching a single SlotMeta is cheap too and if we're calling this function, stuff is kind of messed up already (ie the node is likely going to have to repair the slot)

[AshwinSekar]

In the interest of keeping the backport surface small, i've added a new PurgeType variant, and fetched the slot_meta within the purge function.

In a future master PR I can unify the remaining PurgeType::Exact callers as suggested here #35124 (comment) and remove this variant. tracking in #35281

[carllin]

does it make sense to include purge_special_columns_exact() in here?

fn purge_range() {}
fn purge_range_special_columns(){}
fn do_purge_range(&self, write_batch: &mut WriteBatch, from_slot: Slot, to_slot: Slot, should_purge_special_columns: bool)

[AshwinSekar]

good idea, i just passed through PurgeType instead of separate functions.

[steviez]

Mostly nits, liking the way this is coming together. Will take a final pass shortly

[steviez]

Just a few more nits on comments + simplification. Think we're good for a ship-it after these changes; I can give that quickly after you make the changes but think it'd be nice to get a final pass & approval from Carl too

[steviez]

LGTM, thanks for pushing through after we collectively changed our mind several times. Just to be sure, let's see if we can get @carllin to give one more pass too

[carllin]

nit: I think it's cleaner encapsulation to pass a should_purge_special_columns boolean here, and then have a function PurgeType::should_should_purge_special_columns() -> bool method on PurgeType. This way purge_slot_cleanup_chaining() doesn't have to know about PurgeType at all

[steviez]

This way purge_slot_cleanup_chaining() doesn't have to know about PurgeType at all

I see the merit of adding the new function purge_slot_cleanup_chaining() as this has the unique functionality from run_purge_with_stats. However, I'm not sure I see the benefit of the boolean over the enum; can you elaborate?

The enum is part of the public API, so I think it is reasonable to expect someone to know about it. And the Exact value of the enum means "go purge the special columns for these slots right now", which is what the boolean would convey.

[carllin]

1. It seems like the PurgeType was designed specifically for the run_purge_with_stats workflow, so it seemed better to keep it isolated there rather than mixing it with this new clean function
2. purge_slot_cleanup_chaining as a utility function seems like it should just decide which columns to clean. Seems like cramming the PurgeType in there was too high level.