Hi SimonReach,
Stacking is not stable with CS9300X. I do face the same issue , this is a bug in the firmware.
Previous discussion with Meraki tac:
Once again, thank you for your patience during these past troubleshooting calls. I am writing this email to provide a summary of the various issues that resulted in instability with the C9300 stacks in this network, the changes made to address these issues, and our recommended next steps.
Originally, we observed that ports on the Core stack went into an LACP blocking state when a member of a downstream stack was rebooted. By taking packet captures on the Core stacks, we confirmed that the downstream stacks had stopped sending LACP PDUs following the reboot of their active member, resulting in the Core stack going into an LACP blocking state. Further investigation confirmed that this was due to an internal issue in which a C9300 stack running CS 16.x firmware may not properly reapply its LACP config following the reboot of the active member of the stack.
To address this CS 16.x LACP issue, we tested upgrading the Core stack and one downstream stack to Beta firmware version CS 17.1.2.1, as a fix for that issue is included in this firmware. After performing this upgrade and testing failover again, we observed the LACP instability issue did not persist, however, a separate issue occurred that caused the downstream stack to go offline for 10-20 minutes until the downstream stack automatically rebooted. After examining the boot reason, we confirmed that this behavior was caused by a separate internal issue in which a C9300 stack running CS 17.x firmware does not properly failover when a stack member is rebooted.
Due to the two issues above, the decision was made to revert the firmware to stable release CS 16.8 and remove LACP configs. Once this was completed and the network was stabilized, we again tested the failover by power cycling the active member of a downstream stack once more. After doing so, the failover time was approximately 4 minutes. Due to the necessary RSTP convergence needed for this failover, along with the expected behavior for C9300 stack failover times, this duration would be considered within normal range.
Going forward, we recommend continuing to monitor the network for stability in its current configuration. Once a stable firmware patch is released that includes a fix for both the LACP instability issue in CS 16 and the stack failover instability issue in CS 17, an additional update will be provided on this case.
