Issue with a 'vehicle' network.

Switch 1-------------------------------------------------Sw3------------------------------------Sw5

| | |

| | |

| | |

Unmanaged SW ---- Device 1 Device 2 Device 3

|| | |

|| | |

|| | |

Switch 2-----------------------------------------------Sw4--------------------------------------Sw6

The figure above describes a network we plan to implement on a moving vehicle. The devices represent computers which send signals to the propulsion units on the vehicle allowing it to accelerate, decelerate etc. With the exception of the unmanged switch, all switches are L3 with a single VLAN spanning across the entire network shown above.

The idea is for Device 1 to be able to constantly communicate with either Device 2 or 3 at any given time. Originally Device 1 was only connected to Sw1, but to add more redundancy, a second link ( represented by ||) was proposed. This was supposed to allow for a single failure (of either Sw1 or any of the links) and provide an alternate path for Device 1 to communicate to 2 and 3. During some basic connectivity tests, I ran simultaneous and continuous pings from Device 1 to Devices 2 and 3. However, upon unplugging any of the primary links or killing Sw1, both ping streams are affected and it takes about 30 seconds for them to recover. The transition is not seamless and this 30 second gap is considered an unacceptable failure condition.

All the devices are on the same network with RSTP running on all switches. Trunk links have been established and the behavior is generally predictable. The unmanaged switch however is from a different supplier and is flat with no configuration on it. The ports on Switches 1 and 2 connected to this unmanaged switch are access ports. What is causing this behaviour? What can be done to rectify this? Thanks!

​

EDIT: Sorry, the formatting is screwed up! The crude illustration was that of a ladder type network with Switch 3 and 4 connected to Device 2 and Switch 5 and 6 connected to Device 3.

EDIT2: Grammar.