From: Remi Gacogne <remi.gacogne@powerdns.com>
Date: Thu, 3 Jan 2019 15:14:28 +0000 (+0100)
Subject: dnsdist: Add a 'rise' parameter to 'newServer()'
X-Git-Tag: rec-4.2.0-alpha1~21^2
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=1b633bec2e1ac8d474be63b9313c55bae2d1d738;p=pdns

dnsdist: Add a 'rise' parameter to 'newServer()'
---

diff --git a/pdns/dnsdist-lua.cc b/pdns/dnsdist-lua.cc
index 091e8e505..49bdde30e 100644
--- a/pdns/dnsdist-lua.cc
+++ b/pdns/dnsdist-lua.cc
@@ -368,6 +368,10 @@ void setupLuaConfig(bool client)
 			  ret->maxCheckFailures=std::stoi(boost::get<string>(vars["maxCheckFailures"]));
 			}
 
+                        if(vars.count("rise")) {
+                          ret->minRiseSuccesses=std::stoi(boost::get<string>(vars["rise"]));
+                        }
+
                         if(vars.count("cpus")) {
                           for (const auto cpu : boost::get<vector<pair<int,string>>>(vars["cpus"])) {
                             cpus.insert(std::stoi(cpu.second));
diff --git a/pdns/dnsdist.cc b/pdns/dnsdist.cc
index a6c4e7acc..aafe16ac9 100644
--- a/pdns/dnsdist.cc
+++ b/pdns/dnsdist.cc
@@ -1930,14 +1930,32 @@ static void healthChecksThread()
       if(dss->availability==DownstreamState::Availability::Auto) {
         bool newState=upCheck(*dss);
         if (newState) {
-          if (dss->currentCheckFailures != 0) {
-            dss->currentCheckFailures = 0;
+          /* check succeeded */
+          dss->currentCheckFailures = 0;
+
+          if (!dss->upStatus) {
+            /* we were marked as down */
+            dss->consecutiveSuccesfulChecks++;
+            if (dss->consecutiveSuccesfulChecks < dss->minRiseSuccesses) {
+              /* if we need more than one successful check to rise
+                 and we didn't reach the threshold yet,
+                 let's stay down */
+              newState = false;
+            }
           }
         }
-        else if (!newState && dss->upStatus) {
-          dss->currentCheckFailures++;
-          if (dss->currentCheckFailures < dss->maxCheckFailures) {
-            newState = true;
+        else {
+          /* check failed */
+          dss->consecutiveSuccesfulChecks = 0;
+
+          if (dss->upStatus) {
+            /* we are currently up */
+            dss->currentCheckFailures++;
+            if (dss->currentCheckFailures < dss->maxCheckFailures) {
+              /* we need more than one failure to be marked as down,
+                 and we did not reach the threshold yet, let's stay down */
+              newState = true;
+            }
           }
         }
 
@@ -1954,6 +1972,7 @@ static void healthChecksThread()
 
           dss->upStatus = newState;
           dss->currentCheckFailures = 0;
+          dss->consecutiveSuccesfulChecks = 0;
           if (g_snmpAgent && g_snmpTrapsEnabled) {
             g_snmpAgent->sendBackendStatusChangeTrap(dss);
           }
diff --git a/pdns/dnsdist.hh b/pdns/dnsdist.hh
index bbeeb0bcb..0827bcdba 100644
--- a/pdns/dnsdist.hh
+++ b/pdns/dnsdist.hh
@@ -711,7 +711,9 @@ struct DownstreamState
   uint16_t xpfRRCode{0};
   uint16_t checkTimeout{1000}; /* in milliseconds */
   uint8_t currentCheckFailures{0};
+  uint8_t consecutiveSuccesfulChecks{0};
   uint8_t maxCheckFailures{1};
+  uint8_t minRiseSuccesses{1};
   StopWatch sw;
   set<string> pools;
   enum class Availability { Up, Down, Auto} availability{Availability::Auto};
diff --git a/pdns/dnsdistdist/docs/reference/config.rst b/pdns/dnsdistdist/docs/reference/config.rst
index 6015f7808..193dda46b 100644
--- a/pdns/dnsdistdist/docs/reference/config.rst
+++ b/pdns/dnsdistdist/docs/reference/config.rst
@@ -312,6 +312,7 @@ Servers
 
   .. versionchanged:: 1.3.4
     - Added ``checkTimeout`` to server_table
+    - Added ``rise`` to server_table.
 
   Add a new backend server. Call this function with either a string::
 
@@ -339,8 +340,8 @@ Servers
       checkClass=NUM,        -- Use NUM as QCLASS in the health-check query, default: DNSClass.IN
       checkName=STRING,      -- Use STRING as QNAME in the health-check query, default: "a.root-servers.net."
       checkType=STRING,      -- Use STRING as QTYPE in the health-check query, default: "A"
-      checkFunction=FUNCTION -- Use this function to dynamically set the QNAME, QTYPE and QCLASS to use in the health-check query (see :ref:`Healthcheck`)
-      checkTimeout=NUM       -- The timeout (in milliseconds) of a health-check query, default: 1000 (1s)
+      checkFunction=FUNCTION,-- Use this function to dynamically set the QNAME, QTYPE and QCLASS to use in the health-check query (see :ref:`Healthcheck`)
+      checkTimeout=NUM,      -- The timeout (in milliseconds) of a health-check query, default: 1000 (1s)
       setCD=BOOL,            -- Set the CD (Checking Disabled) flag in the health-check query, default: false
       maxCheckFailures=NUM,  -- Allow NUM check failures before declaring the backend down, default: 1
       mustResolve=BOOL,      -- Set to true when the health check MUST return a NOERROR RCODE and an answer
@@ -353,7 +354,8 @@ Servers
       addXPF=NUM,            -- Add the client's IP address and port to the query, along with the original destination address and port,
                              -- using the experimental XPF record from `draft-bellis-dnsop-xpf <https://datatracker.ietf.org/doc/draft-bellis-dnsop-xpf/>`_ and the specified option code. Default is disabled (0)
       sockets=NUM,           -- Number of sockets (and thus source ports) used toward the backend server, defaults to a single one
-      disableZeroScope       -- Disable the EDNS Client Subnet 'zero scope' feature, which does a cache lookup for an answer valid for all subnets (ECS scope of 0) before adding ECS information to the query and doing the regular lookup
+      disableZeroScope=BOOL, -- Disable the EDNS Client Subnet 'zero scope' feature, which does a cache lookup for an answer valid for all subnets (ECS scope of 0) before adding ECS information to the query and doing the regular lookup
+      rise=NUM               -- Require NUM consecutive successful checks before declaring the backend up, default: 1
     })
 
   :param str server_string: A simple IP:PORT string.