You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This is the meta issue for the fast TXMeta project.
The goal is to decouple Horizon from the stellar-core database, remove the need for stellar-core to maintain txmeta internally, and allow Horizon to come up to date with historical ledger information quickly and easily.
"Offline mode" means the bulk replay mode where a captive stellar-core with in-memory ledger processing is used to generate historical txmeta very quickly.
"Online mode" means the current-ledger trickle mode where a captive stellar-core provides on-going ledger updates in real-time.
"FSC" means "fast stellar-core", a stellar-core instance running in-memory solely for the purpose of producing txmeta.
Produce a benchmark that demonstrates that Horizon can complete a full ingestion in under 24 hours on a x1.32xlarge AWS machine.
Online mode
Discuss whether we need online mode to run in-memory
Identify any needed work on stellar-core side and raise issues
Update interface to communicate bulk (offline) vs trickle (online) mode
Update prototype to support online mode
Produce a test demonstrating code can catch up offline and then switch to online mode
Confirm crash/restart works reliably during online mode (no corruption etc)
Identify how to implement transaction submission
Clustering
Remove requirement for front-end Horizon nodes to directly ingest to memory
Implement support for multiple online Horizon-FSC instances for redundancy
Packaging
Determine how to optimally package and distribute Horizon-FSC for ease of use
Cleanup
Remove Horizon accesses to stellar-core DB during ingestion
Remove stellar-core queries made from the txsub package for the case where Horizon is configured to ignore failed transactions (implement a short lived failed transactions cache in ingestion instead?) and transaction status is requested.
Clean up and remove all cursors and any vestigial code that accesses the stellar-core database.
Stretch goals (nice to have)
Implement XDR contract changes to allow partial ledger processing.
Fall back to sqlite for more modest hardware / very large ledgers
Confirm Horizon-FSC runs on Windows (Docker container sufficient)
The text was updated successfully, but these errors were encountered:
This is the meta issue for the fast TXMeta project.
The goal is to decouple Horizon from the stellar-core database, remove the need for stellar-core to maintain txmeta internally, and allow Horizon to come up to date with historical ledger information quickly and easily.
"Offline mode" means the bulk replay mode where a captive stellar-core with in-memory ledger processing is used to generate historical txmeta very quickly.
"Online mode" means the current-ledger trickle mode where a captive stellar-core provides on-going ledger updates in real-time.
"FSC" means "fast stellar-core", a stellar-core instance running in-memory solely for the purpose of producing txmeta.
Offline mode
Horizon can complete a full ingestion in under 24 hours on a x1.32xlarge AWS machine.
Online mode
Clustering
Packaging
Cleanup
txsub
package for the case where Horizon is configured to ignore failed transactions (implement a short lived failed transactions cache in ingestion instead?) and transaction status is requested.Stretch goals (nice to have)
The text was updated successfully, but these errors were encountered: