F
Flow•16mo ago
Jan

jan.6550 (2023-07-13)

Hey everyone, if you experienced issues with script execution on Testnet this should be resolved now, please let us know if you still have any problems
6 Replies
Needle
Needle•16mo ago
I've created a thread for your message. Please continue any relevant discussion in this thread. You can rename this thread using /title <new title> If this is a technical question that others may benefit from, considering also asking it on Stackoverflow: https://stackoverflow.com/questions/ask?tags=onflow-cadence
Unknown User
Unknown User•16mo ago
Message Not Public
Sign In & Join Server To View
Jan
Jan•16mo ago
Totally, I will post a quick explainer here today 🙂 Quick summary of the testnet issue - please let me know if you want to know more! We are working on moving script execution from execution nodes (ENs) to archive node (RN), because that enables horizontal scalability of script execution and also allows developers to run scripts locally on their instance of RN. RNs store the execution state on disk (in Pebble DB) and so are much cheaper to run then ENs. On Jul 13th we have been testing script execution on Archive nodes. It went well in the sense that the RN with 32GB of memory could take pretty much all of the devnet script load (~8 scripts executed per second on average). The issue was with the data the RN is using to update it's execution state. The data is uploaded to GCP bucket by an Execution node and it looks like for some reason it is not uploading complete data at times, so when Archive node updates it's execution state based on this data in GCP, it is incomplete. The dapps that relied on the scripts executed on the RN started to see this data inconsistency. We are switching the RN to execution data sync in the long run, instead of this GCP upload mechanism, so this can be fixed fairly easily we think. The main purpose of this test was to collect first data on the performance with Pebble DB.
Unknown User
Unknown User•16mo ago
Message Not Public
Sign In & Join Server To View
Jan
Jan•16mo ago
Hey Bjartek, On Archive node specifically Pebble was used instead of Badger for execution state register storage. Badger is not actively supported and our performance benchmark also showed better performance for Pebble. We also plan to replace Badger with Pebble in general on most nodes but have no specific plans yet.
Unknown User
Unknown User•16mo ago
Message Not Public
Sign In & Join Server To View
Want results from more Discord servers?
Add your server