Release 3.9.5

1

api-service: Fixed an issue causing the api-service to become unresponsive in case it loses etcd connectivity during the service's startup.

LBM1-34970

2

cluster-manager: Fixed a bug in volumeFSM that used to flood the logs. Each time the FSM reached the updating state, Lightbits printed to log and invoked the FSM again. The fix assures that we only retrigger once.

LBM1-34501

3

cluster-manager: Modified the handling of volume delete requests so that the critical updates to etcd will be transactional, and will not leave the system with partial data even when the active cluster-manager is switched during the process.

LBM1-34975

4

discovery-service & node-manager: Fixed an issue that could potentially cause volumes to not appear on clients after parallel client reboots and node-manager restarts/failovers.

LBM1-34260

5

duroslight: Fixed a connectivity reporting issue that could cause a node to become inactive (due to a "Connectivity Issue"), when another node either fails or is intentionally shut down (e.g., due to reboot or node-manager restart).

LBM1-35450

6

userlbe: Fixed a rare race when Garbage Collector was reading statistics while Metadata was updating the statistics. This could result in the userlbe process crashing.

LBM1-34613

7

duroslight: Set the target side keep-alive to 10 seconds more than specified in the connect command from the client. This is to prevent spurious keep-alive failures with client kernels that incorrectly use identical keep-alive settings for client and target.

LBM1-35187

8

userlbe: Fixed rare Global Flash Translation Layer (GFTL) assertion failure during multiple rebuilds: MBD_recovery could fail to put a completion due to transient queue full condition.

LBM1-35664