Clients should not wait infinitely for etcd to start if it's unavailable

src/client/cluster_client.cpp

@@ -58,6 +58,7 @@ cluster_client_t::cluster_client_t(ring_loop_t *ringloop, timerfd_manager_t *tfd
     st_cli.on_reload_hook = [this]() { st_cli.load_global_config(); };
 
     st_cli.parse_config(config);
+    st_cli.infinite_start = false;
     st_cli.load_global_config();
 
     scrap_buffer_size = SCRAP_BUFFER_SIZE;

src/client/etcd_state_client.cpp

@@ -121,6 +121,7 @@ void etcd_state_client_t::etcd_call(std::string api, json11::Json payload, int t
         "Connection: keep-alive\r\n"
         "Keep-Alive: timeout="+std::to_string(etcd_keepalive_timeout)+"\r\n"
         "\r\n"+req;
+    retries--;
     auto cb = [this, api, payload, timeout, retries, interval, callback,
         cur_addr = selected_etcd_address](const http_response_t *response)
     {
@@ -144,11 +145,11 @@ void etcd_state_client_t::etcd_call(std::string api, json11::Json payload, int t
                 {
                     tfd->set_timer(interval, false, [this, api, payload, timeout, retries, interval, callback](int)
                     {
-                        etcd_call(api, payload, timeout, retries-1, interval, callback);
+                        etcd_call(api, payload, timeout, retries, interval, callback);
                     });
                 }
                 else
-                    etcd_call(api, payload, timeout, retries-1, interval, callback);
+                    etcd_call(api, payload, timeout, retries, interval, callback);
             }
             else
                 callback(err, data);
@@ -558,15 +559,22 @@ void etcd_state_client_t::load_global_config()
 {
     etcd_call("/kv/range", json11::Json::object {
         { "key", base64_encode(etcd_prefix+"/config/global") }
-    }, etcd_slow_timeout, max_etcd_attempts, 0, [this](std::string err, json11::Json data)
+    }, etcd_quick_timeout, max_etcd_attempts, 0, [this](std::string err, json11::Json data)
     {
         if (err != "")
         {
-            fprintf(stderr, "Error reading OSD configuration from etcd: %s\n", err.c_str());
-            tfd->set_timer(etcd_slow_timeout, false, [this](int timer_id)
+            fprintf(stderr, "Error reading configuration from etcd: %s\n", err.c_str());
+            if (infinite_start)
             {
-                load_global_config();
-            });
+                tfd->set_timer(etcd_slow_timeout, false, [this](int timer_id)
+                {
+                    load_global_config();
+                });
+            }
+            else
+            {
+                exit(1);
+            }
             return;
         }
         json11::Json::object global_config;

src/client/etcd_state_client.h

@@ -106,6 +106,7 @@ public:
     int max_etcd_attempts = 5;
     int etcd_quick_timeout = 1000;
     int etcd_slow_timeout = 5000;
+    bool infinite_start = true;
     uint64_t global_block_size = DEFAULT_BLOCK_SIZE;
     uint32_t global_bitmap_granularity = DEFAULT_BITMAP_GRANULARITY;
     uint32_t global_immediate_commit = IMMEDIATE_NONE;

src/cmd/cli.cpp

@@ -485,8 +485,6 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
         p->ringloop = new ring_loop_t(RINGLOOP_DEFAULT_SIZE);
         p->epmgr = new epoll_manager_t(p->ringloop);
         p->cli = new cluster_client_t(p->ringloop, p->epmgr->tfd, cfg_j);
-        // Smaller timeout by default for more interactiveness
-        p->cli->st_cli.etcd_slow_timeout = p->cli->st_cli.etcd_quick_timeout;
         p->loop_and_wait(action_cb, [&](const cli_result_t & r)
         {
             result = r;