Keep track of the number of build steps that are being built

(As opposed to being in the closure copying stage.)

src/hydra-queue-runner/build-remote.cc

@@ -114,7 +114,7 @@ void buildRemote(std::shared_ptr<StoreAPI> store,
     const string & sshName, const string & sshKey,
     const Path & drvPath, const Derivation & drv,
     const nix::Path & logDir, unsigned int maxSilentTime, unsigned int buildTimeout,
-    RemoteResult & result)
+    RemoteResult & result, counter & nrStepsBuilding)
 {
     string base = baseNameOf(drvPath);
     result.logFile = logDir + "/" + string(base, 0, 2) + "/" + string(base, 2);
@@ -176,7 +176,11 @@ void buildRemote(std::shared_ptr<StoreAPI> store,
     // FIXME: send maxLogSize.
     to.flush();
     result.startTime = time(0);
-    int res = readInt(from);
+    int res;
+    {
+        MaintainCount mc(nrStepsBuilding);
+        res = readInt(from);
+    }
     result.stopTime = time(0);
     if (res) {
         result.errorMsg = (format("%1% on ‘%2%’") % readString(from) % sshName).str();

src/hydra-queue-runner/build-remote.hh

@@ -3,6 +3,8 @@
 #include "store-api.hh"
 #include "derivations.hh"
 
+#include "counter.hh"
+
 struct RemoteResult
 {
     enum {
@@ -20,4 +22,4 @@ void buildRemote(std::shared_ptr<nix::StoreAPI> store,
     const std::string & sshName, const std::string & sshKey,
     const nix::Path & drvPath, const nix::Derivation & drv,
     const nix::Path & logDir, unsigned int maxSilentTime, unsigned int buildTimeout,
-    RemoteResult & result);
+    RemoteResult & result, counter & nrStepsBuilding);

src/hydra-queue-runner/counter.hh

@@ -0,0 +1,12 @@
+#pragma once
+
+#include <atomic>
+
+typedef std::atomic<unsigned int> counter;
+
+struct MaintainCount
+{
+    counter & c;
+    MaintainCount(counter & c) : c(c) { c++; }
+    ~MaintainCount() { c--; }
+};

src/hydra-queue-runner/hydra-queue-runner.cc

@@ -19,6 +19,7 @@
 #include "build-remote.hh"
 #include "sync.hh"
 #include "pool.hh"
+#include "counter.hh"
 
 #include "store-api.hh"
 #include "derivations.hh"
@@ -43,16 +44,6 @@ bool has(const C & c, const V & v)
 }
 
 
-typedef std::atomic<unsigned int> counter;
-
-struct MaintainCount
-{
-    counter & c;
-    MaintainCount(counter & c) : c(c) { c++; }
-    ~MaintainCount() { c--; }
-};
-
-
 typedef enum {
     bsSuccess = 0,
     bsFailed = 1,
@@ -252,6 +243,7 @@ private:
     counter nrBuildsDone{0};
     counter nrStepsDone{0};
     counter nrActiveSteps{0};
+    counter nrStepsBuilding{0};
     counter nrRetries{0};
     counter maxNrRetries{0};
     counter nrQueueWakeups{0};
@@ -1094,7 +1086,7 @@ bool State::doBuildStep(std::shared_ptr<StoreAPI> store, Step::ptr step,
         try {
             /* FIXME: referring builds may have conflicting timeouts. */
             buildRemote(store, machine->sshName, machine->sshKey, step->drvPath, step->drv,
-                logDir, build->maxSilentTime, build->buildTimeout, result);
+                logDir, build->maxSilentTime, build->buildTimeout, result, nrStepsBuilding);
         } catch (Error & e) {
             result.status = RemoteResult::rrMiscFailure;
             result.errorMsg = e.msg();
@@ -1432,6 +1424,7 @@ void State::dumpStatus()
         printMsg(lvlError, format("%1% runnable build steps") % runnable_->size());
     }
     printMsg(lvlError, format("%1% active build steps") % nrActiveSteps);
+    printMsg(lvlError, format("%1% build steps currently building") % nrStepsBuilding);
     printMsg(lvlError, format("%1% builds read from queue") % nrBuildsRead);
     printMsg(lvlError, format("%1% builds done") % nrBuildsDone);
     printMsg(lvlError, format("%1% build steps done") % nrStepsDone);