Internal Validator - sequencer and stake manager error

[BlockgenStudio]

Description:

For the past three weeks, we have been experiencing recurring errors on one of our validators, validator-002, in our production environment. We identified three specific errors in the logs, detailed below:

Error Logs:

1. Panic Error
   panic: runtime error: slice bounds out of range [:32] with capacity 0

   • Frequency: ~10 occurrences per week

2. Failed to Run Sequence Error
   failed to run sequence - validator manager init: height=17657995 error="getting voting power failed - backend is not initialized for height 17657995, fsm height 17657994"

   • Frequency: ~20 occurrences per week

3. Post Block in Stake Manager Error
   polygon.server.polybft.consensus_runtime: failed to post block in stake manager: err="not found"

   • Frequency: Appears on every block sequence

These logs are from validator-002 for the time period from October 28th to November 4th.

How to Reproduce the Issue:
Below are the setup and resource details used to set up our Polygon Supernet, along with relevant environment details:

Infrastructure Setup:

• Total Nodes: 7 Validators, 3 Non-validators on an internal network
• Validator Configuration:
  • 5 Validators in a private subnet (genesis validators)
  • 1 Validator in a public subnet
  • 1 External Validator hosted outside the VPC (connected via an RPC from a publicly exposed RPC node to an internal genesis validator)
• Non-Validator Configuration:
  • 2 Non-validators connected to a load balancer, used as RPC nodes
  • 1 Non-validator connected to a block explorer

Resource Details:

• Validator Instance Type: c6i.large
• Non-validator Instance Type (RPC nodes): c6i.xlarge
• Non-validator Instance Type (Block explorer): c6i.2xlarge
• Operating System: Ubuntu 22.04 LTS
• Polygon Edge Version: v1.0.0

Impact and Urgency:
These errors are impacting our production environment. The failed to run sequence error appears multiple times a week and has potential implications for validator stability. Additionally, the post block in stake manager error affects every block sequence, which is a significant operational concern.

Request for Assistance:
Could you provide any guidance on troubleshooting or potential fixes for these issues? If additional logs or specific configurations are needed, please let us know.
Thank you for your assistance!

[R-Santev]

@BlockgenStudio About point 2: This happens when a block is synced from another validator, and latestHeader.Number is updated just before a new sequence is initiated here:

sequenceCh, stopSequence = p.ibft.runSequence(latestHeader.Number + 1)

This occurs because latestHeader is a pointer. While there may be more elegant solutions to handle this case, I believe the Edge team intentionally left it to throw an error since it avoids adding additional logic (and computation) and works well in practice.

@Stefan-Ethernal, feel free to correct me if I’m wrong.

[BlockgenStudio]

thanks @R-Santev what about the 1. and 3.?

[BlockgenStudio]

especially the 3.