diff --git a/src/nfs/nfs_proxy.cpp b/src/nfs/nfs_proxy.cpp
index 30d46071..d5e173b2 100644
--- a/src/nfs/nfs_proxy.cpp
+++ b/src/nfs/nfs_proxy.cpp
@@ -522,6 +522,7 @@ void nfs_proxy_t::do_accept(int listen_fd)
         {
             cli->proc_table.insert(fn);
         }
+        rpc_clients[nfs_fd] = cli;
         epmgr->tfd->set_fd_handler(nfs_fd, true, [cli](int nfs_fd, int epoll_events)
         {
             // Handle incoming event
@@ -531,7 +532,6 @@ void nfs_proxy_t::do_accept(int listen_fd)
                 if (parent->trace)
                     fprintf(stderr, "Client %d disconnected\n", nfs_fd);
                 cli->stop();
-                parent->active_connections--;
                 parent->check_exit();
                 return;
             }
@@ -776,6 +776,8 @@ bool nfs_client_t::deref()
 
 void nfs_client_t::stop()
 {
+    parent->rpc_clients.erase(nfs_fd);
+    parent->active_connections--;
     stopped = true;
     if (refs <= 0)
     {
@@ -1056,6 +1058,11 @@ int nfs_client_t::handle_rpc_message(void *base_buf, void *msg_buf, uint32_t msg
 
 void nfs_proxy_t::daemonize()
 {
+    // Stop all clients because client I/O sometimes breaks during daemonize
+    // I.e. the new process stops receiving events on the old FD
+    // It doesn't happen if we call sleep(1) here, but we don't want to call sleep(1)...
+    for (auto & clp: rpc_clients)
+        clp.second->stop();
     if (fork())
         exit(0);
     setsid();
diff --git a/src/nfs/nfs_proxy.h b/src/nfs/nfs_proxy.h
index f6aa9f3b..29f12019 100644
--- a/src/nfs/nfs_proxy.h
+++ b/src/nfs/nfs_proxy.h
@@ -21,6 +21,7 @@ class cli_tool_t;
 
 struct kv_fs_state_t;
 struct block_fs_state_t;
+class nfs_client_t;
 
 class nfs_proxy_t
 {
@@ -54,6 +55,7 @@ public:
     vitastorkv_dbw_t *db = NULL;
     kv_fs_state_t *kvfs = NULL;
     block_fs_state_t *blockfs = NULL;
+    std::map<int, nfs_client_t*> rpc_clients;
 
     std::vector<XDR*> xdr_pool;