From 7a0b5212feab2fccc58162fb70521967e89cc528 Mon Sep 17 00:00:00 2001
From: Vitaliy Filippov <vitalif@yourcmc.ru>
Date: Sun, 28 Nov 2021 01:43:31 +0300
Subject: [PATCH] Exit if unable to restart watches

FIXME: It's probably not OK for the client to exit in this case
---
 mon/mon.js                | 25 +++++++++++++++++++------
 src/etcd_state_client.cpp | 14 ++++++++++++++
 2 files changed, 33 insertions(+), 6 deletions(-)

diff --git a/mon/mon.js b/mon/mon.js
index cbc9c7da..95bee44a 100644
--- a/mon/mon.js
+++ b/mon/mon.js
@@ -341,7 +341,7 @@ class Mon
         this.etcd_start_timeout = (config.etcd_start_timeout || 5) * 1000;
         this.state = JSON.parse(JSON.stringify(this.constructor.etcd_tree));
         this.signals_set = false;
-        this.on_stop_cb = () => this.on_stop().catch(console.error);
+        this.on_stop_cb = () => this.on_stop(0).catch(console.error);
     }
 
     parse_etcd_addresses(addrs)
@@ -530,12 +530,25 @@ class Mon
             catch (e)
             {
             }
-            if (!data || !data.result || !data.result.events)
+            if (!data || !data.result)
             {
-                if (!data || !data.result || !data.result.watch_id)
+                console.error('Unknown message received from watch websocket: '+msg);
+            }
+            else if (data.result.canceled)
+            {
+                // etcd watch canceled
+                if (data.result.compact_revision)
                 {
-                    console.error('Garbage received from watch websocket: '+msg);
+                    // we may miss events if we proceed
+                    console.error('Revisions before '+data.result.compact_revision+' were compacted by etcd, exiting');
+                    this.on_stop(1);
                 }
+                console.error('Watch canceled by etcd, reason: '+data.result.cancel_reason+', exiting');
+                this.on_stop(1);
+            }
+            else if (data.result.created)
+            {
+                // etcd watch created
             }
             else
             {
@@ -639,11 +652,11 @@ class Mon
         }
     }
 
-    async on_stop()
+    async on_stop(status)
     {
         clearInterval(this.lease_timer);
         await this.etcd_call('/lease/revoke', { ID: this.etcd_lease_id }, this.config.etcd_mon_timeout, this.config.etcd_mon_retries);
-        process.exit(0);
+        process.exit(status);
     }
 
     async become_master()
diff --git a/src/etcd_state_client.cpp b/src/etcd_state_client.cpp
index 3a26eb38..b16ff97b 100644
--- a/src/etcd_state_client.cpp
+++ b/src/etcd_state_client.cpp
@@ -209,6 +209,20 @@ void etcd_state_client_t::start_etcd_watcher()
                 {
                     etcd_watches_initialised++;
                 }
+                if (data["result"]["canceled"].bool_value())
+                {
+                    // etcd watch canceled, maybe because the revision was compacted
+                    if (data["result"]["compact_revision"].uint64_value())
+                    {
+                        // we may miss events if we proceed
+                        // FIXME: reload state and continue when used inside cluster_client
+                        fprintf(stderr, "Revisions before %lu were compacted by etcd, exiting\n",
+                            data["result"]["compact_revision"].uint64_value());
+                        exit(1);
+                    }
+                    fprintf(stderr, "Watch canceled by etcd, reason: %s, exiting\n", data["result"]["cancel_reason"].string_value().c_str());
+                    exit(1);
+                }
                 if (etcd_watches_initialised == 4)
                 {
                     etcd_watch_revision = data["result"]["header"]["revision"].uint64_value();