SEQUENCER CONNECTION POOL BUG IN SPLICE 0.5.9 - 0.5.11

1. #1Wayne Collier20-02-2026source ↗
    

   • Some Validators have had trouble upgrading to Splice 0.5.9 on MainNet due to an issue that causes connections from Validators to sequencers to time out.
     • This issue appears to have shown up only on MainNet, even though Splice 0.5.9 has been available previously on DevNet and TestNet on the normal cadence.
   • This failure is caused by a bug that appears when there's a network outage, or when a network component–like a load balancer–silently drops idle connections.
   • The sequencer connection pool logic should set a "keep alive" that protects against this failure, but in Splice 0.5.9 - Splice 0.5.11 this keep alive was not implemented correctly.
   • The fix for this is in place in Splice 0.5.12, which will once again explicitly configure keep alive in the sequencer connection pool.
   • Splice 0.5.12 is available on DevNet on Monday February 23rd, and will roll out to TestNet and MainNet in the normal two-week sequence.
   • If you've had problems spinning up your Validator on Splice 0.5.9, this could be the cause.
   • You can remove this as a possible issue by disabling the sequencer connection pools in Splice 0.5.9, 0.5.10 and 0.5.11, using the following configuration:
     • Add an environment variable to your participant: ADDITIONAL_CONFIG_DISABLE_NEW_SEQUENCER_CONNECTION_POOL
     • Set the value canton.participants.participant.sequencer-client.use-new-connection-pool=false for this environment variable.

   
   
   

   • Any Validator operator that provides a wallet or application to outside customers should upgrade their Validator nodes on DevNet and TestNet to increase the chance that you'll catch issues before you upgrade your MainNet nodes.
     • DevNet and TestNet exist primarily to allow SVs and Validators to test upgrades.