clean up owned assets involving offline participants #212
+379
−30
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
saketh-are
commented
Feb 14, 2025
| @@ -307,8 +423,10 @@ where | |||
| /// owning some of the assets. Each asset has exactly one owner. | |||
| /// | |||
| /// Only the owner of an asset may pick the asset for use in an MPC computation. | |||
| /// As the owner, the `take_owned` method removes the oldest asset from the | |||
There was a problem hiding this comment.
Even before this PR, we did not actually guarantee that we picked the oldest usable asset. This will likely need to be revisited in the future if we implement a strategy for agreement between nodes on their available assets. For now, I don't believe anything is sensitive to the asset ordering once they are generated.
|
|
||
| if !should_generate { | ||
| // After the store is full, slowly discard triples which cannot be used right now | ||
| triple_store.maybe_discard_owned(10).await; |
There was a problem hiding this comment.
This should be set relative to the buffer size such that recovery happens in a reasonable timeframe. At 10 per .1s, a million buffered triples would be turned over in about 3 hours. Not sure if this is something we'd want to expose in the config.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Motivation
If a node wipes its storage for any reason, any previously generated assets whose participant set includes that node are unusable. Such assets owned by other nodes are stored indefinitely until an attempt is made to use them, leading to failed computations. We need a recovery strategy to address this scenario.
Implementation
Once a node has filled its store with the desired quantity of an asset, it will gradually verify its stored assets and discard those that depend on offline participants.
To recover the system, the wiped node would intentionally remain offline for some time. The other nodes would notice its absence and discard any owned assets that depend on it.
Visibility
Each owned asset type is tracked using three counters:
available,online, andoffline.When the alive participant set changes, both
onlineandofflineare reset to0, indicating that none of the stored assets have been checked yet against the new participant set.If the participant set then remains stable, the stored assets will be verified over time. When
availableandonlineare equal, andofflineis 0, we will know that all assets depending on the offline node have been found and discarded.Closes #213. See also #207.
A time-based expiration on all assets in the DistributedAssetStore will be implemented separately to prevent unbounded accumulation of unowned assets.