hydra-eval-jobs: Parallelize

src/hydra-eval-jobs/hydra-eval-jobs.cc

@@ -1,37 +1,71 @@
 #include <map>
 #include <iostream>
 
-#define GC_LINUX_THREADS 1
-#include <gc/gc_allocator.h>
-
 #include "shared.hh"
 #include "store-api.hh"
 #include "eval.hh"
 #include "eval-inline.hh"
 #include "util.hh"
-#include "json.hh"
 #include "get-drvs.hh"
 #include "globals.hh"
 #include "common-eval-args.hh"
 #include "flake/flakeref.hh"
 #include "flake/flake.hh"
+#include "attr-path.hh"
+#include "derivations.hh"
 
 #include "hydra-config.hh"
 
 #include <sys/types.h>
 #include <sys/wait.h>
+#include <sys/resource.h>
+
+#include <nlohmann/json.hpp>
 
 using namespace nix;
 
-
 static Path gcRootsDir;
+static size_t maxMemorySize;
 
+struct MyArgs : MixEvalArgs, MixCommonArgs
+{
+    Path releaseExpr;
+    bool flake = false;
+    bool dryRun = false;
 
-static void findJobs(EvalState & state, JSONObject & top,
+    MyArgs() : MixCommonArgs("hydra-eval-jobs")
-    Bindings & autoArgs, Value & v, const string & attrPath);
+    {
+        mkFlag()
+            .longName("help")
+            .description("show usage information")
+            .handler([&]() {
+                printHelp(programName, std::cout);
+                throw Exit();
+            });
 
+        mkFlag()
+            .longName("gc-roots-dir")
+            .description("garbage collector roots directory")
+            .labels({"path"})
+            .dest(&gcRootsDir);
 
-static string queryMetaStrings(EvalState & state, DrvInfo & drv, const string & name, const string & subAttribute)
+        mkFlag()
+            .longName("dry-run")
+            .description("don't create store derivations")
+            .set(&dryRun, true);
+
+        mkFlag()
+            .longName("flake")
+            .description("build a flake")
+            .set(&flake, true);
+
+        expectArg("expr", &releaseExpr);
+    }
+};
+
+static MyArgs myArgs;
+
+static std::string queryMetaStrings(EvalState & state, DrvInfo & drv, const string & name, const string & subAttribute)
 {
     Strings res;
     std::function<void(Value & v)> rec;
@@ -56,116 +90,163 @@ static string queryMetaStrings(EvalState & state, DrvInfo & drv, const string &
     return concatStringsSep(", ", res);
 }
 
-
+static void worker(
-static void findJobsWrapped(EvalState & state, JSONObject & top,
+    EvalState & state,
-    Bindings & autoArgs, Value & vIn, const string & attrPath)
+    Bindings & autoArgs,
+    AutoCloseFD & to,
+    AutoCloseFD & from)
 {
-    debug(format("at path `%1%'") % attrPath);
+    Value vTop;
 
-    checkInterrupt();
+    if (myArgs.flake) {
+        using namespace flake;
 
-    Value v;
+        auto flakeRef = parseFlakeRef(myArgs.releaseExpr);
-    state.autoCallFunction(autoArgs, vIn, v);
 
-    if (v.type == tAttrs) {
+        auto vFlake = state.allocValue();
 
-        auto drv = getDerivation(state, v, false);
+        auto lockedFlake = lockFlake(state, flakeRef,
+            LockFlags {
+                .updateLockFile = false,
+                .useRegistries = false,
+                .allowMutable = false,
+            });
 
-        if (drv) {
+        callFlake(state, lockedFlake, *vFlake);
-            Path drvPath;
 
-            DrvInfo::Outputs outputs = drv->queryOutputs();
+        auto vOutputs = vFlake->attrs->get(state.symbols.create("outputs"))->value;
+        state.forceValue(*vOutputs);
 
-            if (drv->querySystem() == "unknown")
+        auto aHydraJobs = vOutputs->attrs->get(state.symbols.create("hydraJobs"));
-                throw EvalError("derivation must have a ‘system’ attribute");
+        if (!aHydraJobs)
+            aHydraJobs = vOutputs->attrs->get(state.symbols.create("checks"));
+        if (!aHydraJobs)
+            throw Error("flake '%s' does not provide any Hydra jobs or checks", flakeRef);
 
-            {
+        vTop = *aHydraJobs->value;
-            auto res = top.object(attrPath);
-            res.attr("nixName", drv->queryName());
-            res.attr("system", drv->querySystem());
-            res.attr("drvPath", drvPath = drv->queryDrvPath());
-            res.attr("description", drv->queryMetaString("description"));
-            res.attr("license", queryMetaStrings(state, *drv, "license", "shortName"));
-            res.attr("homepage", drv->queryMetaString("homepage"));
-            res.attr("maintainers", queryMetaStrings(state, *drv, "maintainers", "email"));
-            res.attr("schedulingPriority", drv->queryMetaInt("schedulingPriority", 100));
-            res.attr("timeout", drv->queryMetaInt("timeout", 36000));
-            res.attr("maxSilent", drv->queryMetaInt("maxSilent", 7200));
-            res.attr("isChannel", drv->queryMetaBool("isHydraChannel", false));
 
-            /* If this is an aggregate, then get its constituents. */
+    } else {
-            Bindings::iterator a = v.attrs->find(state.symbols.create("_hydraAggregate"));
+        state.evalFile(lookupFileArg(state, myArgs.releaseExpr), vTop);
-            if (a != v.attrs->end() && state.forceBool(*a->value, *a->pos)) {
+    }
-                Bindings::iterator a = v.attrs->find(state.symbols.create("constituents"));
+
-                if (a == v.attrs->end())
+    auto vRoot = state.allocValue();
-                    throw EvalError("derivation must have a ‘constituents’ attribute");
+    state.autoCallFunction(autoArgs, vTop, *vRoot);
-                PathSet context;
+
-                state.coerceToString(*a->pos, *a->value, context, true, false);
+    while (true) {
-                PathSet drvs;
+        /* Wait for the master to send us a job name. */
-                for (auto & i : context)
+        writeLine(to.get(), "next");
-                    if (i.at(0) == '!') {
+
-                        size_t index = i.find("!", 1);
+        auto s = readLine(from.get());
-                        drvs.insert(string(i, index + 1));
+        if (s == "exit") break;
+        if (!hasPrefix(s, "do ")) abort();
+        std::string attrPath(s, 3);
+
+        debug("worker process %d at '%s'", getpid(), attrPath);
+
+        /* Evaluate it and send info back to the master. */
+        nlohmann::json reply;
+
+        try {
+            auto v = findAlongAttrPath(state, attrPath, autoArgs, *vRoot).first;
+
+            state.forceValue(*v);
+
+            if (auto drv = getDerivation(state, *v, false)) {
+
+                DrvInfo::Outputs outputs = drv->queryOutputs();
+
+                if (drv->querySystem() == "unknown")
+                    throw EvalError("derivation must have a 'system' attribute");
+
+                auto drvPath = drv->queryDrvPath();
+
+                nlohmann::json job;
+
+                job["nixName"] = drv->queryName();
+                job["system"] =drv->querySystem();
+                job["drvPath"] = drvPath;
+                job["description"] = drv->queryMetaString("description");
+                job["license"] = queryMetaStrings(state, *drv, "license", "shortName");
+                job["homepage"] = drv->queryMetaString("homepage");
+                job["maintainers"] = queryMetaStrings(state, *drv, "maintainers", "email");
+                job["schedulingPriority"] = drv->queryMetaInt("schedulingPriority", 100);
+                job["timeout"] = drv->queryMetaInt("timeout", 36000);
+                job["maxSilent"] = drv->queryMetaInt("maxSilent", 7200);
+                job["isChannel"] = drv->queryMetaBool("isHydraChannel", false);
+
+                /* If this is an aggregate, then get its constituents. */
+                auto a = v->attrs->get(state.symbols.create("_hydraAggregate"));
+                if (a && state.forceBool(*a->value, *a->pos)) {
+                    auto a = v->attrs->get(state.symbols.create("constituents"));
+                    if (!a)
+                        throw EvalError("derivation must have a ‘constituents’ attribute");
+
+
+                    PathSet context;
+                    state.coerceToString(*a->pos, *a->value, context, true, false);
+                    for (auto & i : context)
+                        if (i.at(0) == '!') {
+                            size_t index = i.find("!", 1);
+                            job["constituents"].push_back(string(i, index + 1));
+                        }
+
+                    state.forceList(*a->value, *a->pos);
+                    for (unsigned int n = 0; n < a->value->listSize(); ++n) {
+                        auto v = a->value->listElems()[n];
+                        state.forceValue(*v);
+                        if (v->type == tString)
+                            job["namedConstituents"].push_back(state.forceStringNoCtx(*v));
                     }
-                res.attr("constituents", concatStringsSep(" ", drvs));
+                }
+
+                /* Register the derivation as a GC root.  !!! This
+                   registers roots for jobs that we may have already
+                   done. */
+                auto localStore = state.store.dynamic_pointer_cast<LocalFSStore>();
+                if (gcRootsDir != "" && localStore) {
+                    Path root = gcRootsDir + "/" + std::string(baseNameOf(drvPath));
+                    if (!pathExists(root))
+                        localStore->addPermRoot(localStore->parseStorePath(drvPath), root, false);
+                }
+
+                nlohmann::json out;
+                for (auto & j : outputs)
+                    out[j.first] = j.second;
+                job["outputs"] = std::move(out);
+
+                reply["job"] = std::move(job);
             }
 
-            /* Register the derivation as a GC root.  !!! This
+            else if (v->type == tAttrs) {
-               registers roots for jobs that we may have already
+                auto attrs = nlohmann::json::array();
-               done. */
+                StringSet ss;
-            auto localStore = state.store.dynamic_pointer_cast<LocalFSStore>();
+                for (auto & i : v->attrs->lexicographicOrder()) {
-            if (gcRootsDir != "" && localStore) {
-                Path root = gcRootsDir + "/" + std::string(baseNameOf(drvPath));
-                if (!pathExists(root))
-                    localStore->addPermRoot(localStore->parseStorePath(drvPath), root, false);
-            }
-
-            auto res2 = res.object("outputs");
-            for (auto & j : outputs)
-                res2.attr(j.first, j.second);
-
-            }
-        }
-
-        else {
-            if (!state.isDerivation(v)) {
-                for (auto & i : v.attrs->lexicographicOrder()) {
                     std::string name(i->name);
-
+                    if (name.find('.') != std::string::npos || name.find(' ') != std::string::npos) {
-                    /* Skip jobs with dots in the name. */
-                    if (name.find('.') != std::string::npos) {
                         printError("skipping job with illegal name '%s'", name);
                         continue;
                     }
-
+                    attrs.push_back(name);
-                    findJobs(state, top, autoArgs, *i->value,
-                        (attrPath.empty() ? "" : attrPath + ".") + name);
                 }
+                reply["attrs"] = std::move(attrs);
             }
+
+        } catch (EvalError & e) {
+            reply["error"] = filterANSIEscapes(e.msg(), true);
         }
+
+        writeLine(to.get(), reply.dump());
+
+        /* If our RSS exceeds the maximum, exit. The master will
+           start a new process. */
+        struct rusage r;
+        getrusage(RUSAGE_SELF, &r);
+        if ((size_t) r.ru_maxrss > maxMemorySize * 1024) break;
     }
 
-    else if (v.type == tNull) {
+    writeLine(to.get(), "restart");
-        // allow null values, meaning 'do nothing'
-    }
-
-    else
-        throw TypeError(format("unsupported value: %1%") % v);
 }
 
-
-static void findJobs(EvalState & state, JSONObject & top,
-    Bindings & autoArgs, Value & v, const string & attrPath)
-{
-    try {
-        findJobsWrapped(state, top, autoArgs, v, attrPath);
-    } catch (EvalError & e) {
-        auto res = top.object(attrPath);
-        res.attr("error", filterANSIEscapes(e.msg(), true));
-    }
-}
-
-
 int main(int argc, char * * argv)
 {
     /* Prevent undeclared dependencies in the evaluation via
@@ -176,54 +257,14 @@ int main(int argc, char * * argv)
 
         auto config = std::make_unique<::Config>();
 
-        auto initialHeapSize = config->getStrOption("evaluator_initial_heap_size", "");
+        auto nrWorkers = config->getIntOption("evaluator_workers", 1);
-        if (initialHeapSize != "")
+        maxMemorySize = config->getIntOption("evaluator_max_memory_size", 4096);
-            setenv("GC_INITIAL_HEAP_SIZE", initialHeapSize.c_str(), 1);
 
         initNix();
         initGC();
 
-        struct MyArgs : MixEvalArgs, MixCommonArgs
-        {
-            Path releaseExpr;
-            bool flake = false;
-
-            MyArgs() : MixCommonArgs("hydra-eval-jobs")
-            {
-                mkFlag()
-                    .longName("help")
-                    .description("show usage information")
-                    .handler([&]() {
-                        printHelp(programName, std::cout);
-                        throw Exit();
-                    });
-
-                mkFlag()
-                    .longName("gc-roots-dir")
-                    .description("garbage collector roots directory")
-                    .labels({"path"})
-                    .dest(&gcRootsDir);
-
-                mkFlag()
-                    .longName("dry-run")
-                    .description("don't create store derivations")
-                    .set(&settings.readOnlyMode, true);
-
-                mkFlag()
-                    .longName("flake")
-                    .description("build a flake")
-                    .set(&flake, true);
-
-                expectArg("expr", &releaseExpr);
-            }
-        };
-
-        MyArgs myArgs;
         myArgs.parseCmdline(argvToStrings(argc, argv));
 
-        JSONObject json(std::cout, true);
-        std::cout.flush();
-
         /* FIXME: The build hook in conjunction with import-from-derivation is causing "unexpected EOF" during eval */
         settings.builders = "";
 
@@ -235,47 +276,196 @@ int main(int argc, char * * argv)
            'getEnv', 'currentSystem' etc. */
         evalSettings.pureEval = myArgs.flake;
 
+        if (myArgs.dryRun) settings.readOnlyMode = true;
+
         if (myArgs.releaseExpr == "") throw UsageError("no expression specified");
 
         if (gcRootsDir == "") printMsg(lvlError, "warning: `--gc-roots-dir' not specified");
 
-        EvalState state(myArgs.searchPath, openStore());
+        struct State
+        {
+            std::set<std::string> todo{""};
+            std::set<std::string> active;
+            nlohmann::json jobs;
+            std::exception_ptr exc;
+        };
 
-        Bindings & autoArgs = *myArgs.getAutoArgs(state);
+        std::condition_variable wakeup;
 
-        Value v;
+        Sync<State> state_;
 
-        if (myArgs.flake) {
+        /* Start a handler thread per worker process. */
-            using namespace flake;
+        auto handler = [&]()
+        {
+            try {
+                pid_t pid = -1;
+                AutoCloseFD from, to;
 
-            auto flakeRef = parseFlakeRef(myArgs.releaseExpr);
+                while (true) {
 
-            auto vFlake = state.allocValue();
+                    /* Start a new worker process if necessary. */
+                    if (pid == -1) {
+                        Pipe toPipe, fromPipe;
+                        toPipe.create();
+                        fromPipe.create();
+                        pid = startProcess(
+                            [&,
+                             to{std::make_shared<AutoCloseFD>(std::move(fromPipe.writeSide))},
+                             from{std::make_shared<AutoCloseFD>(std::move(toPipe.readSide))}
+                            ]()
+                            {
+                                try {
+                                    EvalState state(myArgs.searchPath, openStore());
+                                    Bindings & autoArgs = *myArgs.getAutoArgs(state);
+                                    worker(state, autoArgs, *to, *from);
+                                } catch (std::exception & e) {
+                                    nlohmann::json err;
+                                    err["error"] = e.what();
+                                    writeLine(to->get(), err.dump());
+                                }
+                            },
+                            ProcessOptions { .allowVfork = false });
+                        from = std::move(fromPipe.readSide);
+                        to = std::move(toPipe.writeSide);
+                        debug("created worker process %d", pid);
+                    }
 
-            auto lockedFlake = lockFlake(state, flakeRef,
+                    /* Check whether the existing worker process is still there. */
-                LockFlags {
+                    auto s = readLine(from.get());
-                    .updateLockFile = false,
+                    if (s == "restart") {
-                    .useRegistries = false,
+                        pid = -1;
-                    .allowMutable = false,
+                        continue;
-                });
+                    } else if (s != "next") {
+                        auto json = nlohmann::json::parse(s);
+                        throw Error("worker error: %s", (std::string) json["error"]);
+                    }
 
-            callFlake(state, lockedFlake, *vFlake);
+                    /* Wait for a job name to become available. */
+                    std::string attrPath;
 
-            auto vOutputs = (*vFlake->attrs->get(state.symbols.create("outputs")))->value;
+                    while (true) {
-            state.forceValue(*vOutputs);
+                        checkInterrupt();
+                        auto state(state_.lock());
+                        if ((state->todo.empty() && state->active.empty()) || state->exc) {
+                            writeLine(to.get(), "exit");
+                            return;
+                        }
+                        if (!state->todo.empty()) {
+                            attrPath = *state->todo.begin();
+                            state->todo.erase(state->todo.begin());
+                            state->active.insert(attrPath);
+                            break;
+                        } else
+                            state.wait(wakeup);
+                    }
 
-            auto aHydraJobs = vOutputs->attrs->get(state.symbols.create("hydraJobs"));
+                    /* Tell the worker to evaluate it. */
-            if (!aHydraJobs)
+                    writeLine(to.get(), "do " + attrPath);
-                aHydraJobs = vOutputs->attrs->get(state.symbols.create("checks"));
-            if (!aHydraJobs)
-                throw Error("flake '%s' does not provide any Hydra jobs or checks", flakeRef);
 
-            v = *(*aHydraJobs)->value;
+                    /* Wait for the response. */
+                    auto response = nlohmann::json::parse(readLine(from.get()));
 
-        } else {
+                    /* Handle the response. */
-            state.evalFile(lookupFileArg(state, myArgs.releaseExpr), v);
+                    StringSet newAttrs;
+
+                    if (response.find("job") != response.end()) {
+                        auto state(state_.lock());
+                        state->jobs[attrPath] = response["job"];
+                    }
+
+                    if (response.find("attrs") != response.end()) {
+                        for (auto & i : response["attrs"]) {
+                            auto s = (attrPath.empty() ? "" : attrPath + ".") + (std::string) i;
+                            newAttrs.insert(s);
+                        }
+                    }
+
+                    if (response.find("error") != response.end()) {
+                        auto state(state_.lock());
+                        state->jobs[attrPath]["error"] = response["error"];
+                    }
+
+                    /* Add newly discovered job names to the queue. */
+                    {
+                        auto state(state_.lock());
+                        state->active.erase(attrPath);
+                        for (auto & s : newAttrs)
+                            state->todo.insert(s);
+                        wakeup.notify_all();
+                    }
+                }
+            } catch (...) {
+                auto state(state_.lock());
+                state->exc = std::current_exception();
+                wakeup.notify_all();
+            }
+        };
+
+        std::vector<std::thread> threads;
+        for (size_t i = 0; i < nrWorkers; i++)
+            threads.emplace_back(std::thread(handler));
+
+        for (auto & thread : threads)
+            thread.join();
+
+        auto state(state_.lock());
+
+        if (state->exc)
+            std::rethrow_exception(state->exc);
+
+        /* For aggregate jobs that have named consistuents
+           (i.e. constituents that are a job name rather than a
+           derivation), look up the referenced job and add it to the
+           dependencies of the aggregate derivation. */
+        auto store = openStore();
+
+        for (auto i = state->jobs.begin(); i != state->jobs.end(); ++i) {
+            auto jobName = i.key();
+            auto & job = i.value();
+
+            auto named = job.find("namedConstituents");
+            if (named == job.end()) continue;
+
+            if (myArgs.dryRun) {
+                for (std::string jobName2 : *named) {
+                    auto job2 = state->jobs.find(jobName2);
+                    if (job2 == state->jobs.end())
+                        throw Error("aggregate job '%s' references non-existent job '%s'", jobName, jobName2);
+                    std::string drvPath2 = (*job2)["drvPath"];
+                    job["constituents"].push_back(drvPath2);
+                }
+            } else {
+                std::string drvPath = job["drvPath"];
+                auto drv = readDerivation(*store, drvPath);
+
+                for (std::string jobName2 : *named) {
+                    auto job2 = state->jobs.find(jobName2);
+                    if (job2 == state->jobs.end())
+                        throw Error("aggregate job '%s' references non-existent job '%s'", jobName, jobName2);
+                    std::string drvPath2 = (*job2)["drvPath"];
+                    auto drv2 = readDerivation(*store, drvPath2);
+                    job["constituents"].push_back(drvPath2);
+                    drv.inputDrvs[store->parseStorePath(drvPath2)] = {drv2.outputs.begin()->first};
+                }
+
+                std::string drvName(store->parseStorePath(drvPath).name());
+                assert(hasSuffix(drvName, drvExtension));
+                drvName.resize(drvName.size() - drvExtension.size());
+                auto h = hashDerivationModulo(*store, drv, true);
+                auto outPath = store->makeOutputPath("out", h, drvName);
+                drv.env["out"] = store->printStorePath(outPath);
+                drv.outputs.insert_or_assign("out", DerivationOutput(outPath.clone(), "", ""));
+                auto newDrvPath = store->printStorePath(writeDerivation(store, drv, drvName));
+
+                debug("rewrote aggregate derivation %s -> %s", drvPath, newDrvPath);
+
+                job["drvPath"] = newDrvPath;
+                job["outputs"]["out"] = store->printStorePath(outPath);
+            }
+
+            job.erase("namedConstituents");
         }
 
-        findJobs(state, json, autoArgs, v, "");
+        std::cout << state->jobs.dump(2) << "\n";
     });
 }