In　a　practical　blockchain　system　based　on　the　Practical　Byzantine　Fault　Tolerance　(PBFT)　protocol,　the　voting　nodes　can　fail　at　any　time　due　to　non-Byzantine　errors,　such　as　autonomous　shutdowns,　device　crashes,　and　communication　link　failures　caused　by　mobility　or　obstacles.　These　errors　may　cause　voting　nodes　to　exit　the　PBFT-based　blockchain　system　unpredictably,　resulting　in　a　variable　number　of　voting　nodes　available　at　any　given　time.　To　maintain　optimal　performance　and　consistency　while　adapting　a　PBFT-based　blockchain　system　to　this　dynamic　change,　this　paper　proposes　an　extension　to　the　PBFT　protocol　by　introducing　a　repair　process　for　failed　nodes.　The　new　PBFT-based　blockchain　system　with　repairable　voting　nodes　is　then　analyzed　for　performance　and　reliability　analysis　by　using　multi-dimensional　Markov　processes,　queueing　theory,　and　the　first　passage　time　method.　Additionally,　we　validate　the　accuracy　of　our　theoretical　findings　by　conducting　numerical　examples　and　simulation　experiments.　These　experiments　demonstrate　that　the　introduction　of　a　repair　process　can　improve　the　performance　and　reliability　of　the　PBFT-based　blockchain　system.　Furthermore,　we　illustrate　how　various　system　parameters　impact　the　performance　measures　of　the　PBFT-based　blockchain　system　with　repairable　voting　nodes.　We　hope　that　the　methodology　and　results　presented　in　this　paper　will　establish　a　common　framework　for　deriving　theoretical　analysis　of　existing　PBFT-based　blockchain　systems　and　inspire　future　research　efforts　in　this　field.　IEEE