lix/src/build-remote/build-remote.cc

303 lines
10 KiB
C++
Raw Normal View History

2016-07-18 22:50:27 +00:00
#include <cstdlib>
#include <cstring>
#include <algorithm>
#include <set>
2016-07-18 22:50:27 +00:00
#include <memory>
#include <tuple>
#include <iomanip>
2017-01-24 12:57:26 +00:00
#if __APPLE__
#include <sys/time.h>
#endif
2016-07-18 22:50:27 +00:00
#include "shared.hh"
#include "pathlocks.hh"
#include "globals.hh"
#include "serialise.hh"
#include "store-api.hh"
#include "derivations.hh"
using namespace nix;
using std::cin;
2017-03-03 15:18:49 +00:00
static void handleAlarm(int sig) {
2016-07-18 22:50:27 +00:00
}
2017-03-03 15:18:49 +00:00
class Machine {
const std::set<string> supportedFeatures;
const std::set<string> mandatoryFeatures;
2016-07-18 22:50:27 +00:00
public:
const string storeUri;
2016-07-18 22:50:27 +00:00
const std::vector<string> systemTypes;
const string sshKey;
2017-03-03 15:18:49 +00:00
const unsigned int maxJobs;
const unsigned int speedFactor;
bool enabled = true;
2016-07-18 22:50:27 +00:00
bool allSupported(const std::set<string> & features) const {
2016-07-18 22:50:27 +00:00
return std::all_of(features.begin(), features.end(),
[&](const string & feature) {
return supportedFeatures.count(feature) ||
mandatoryFeatures.count(feature);
2016-07-18 22:50:27 +00:00
});
}
bool mandatoryMet(const std::set<string> & features) const {
2016-07-18 22:50:27 +00:00
return std::all_of(mandatoryFeatures.begin(), mandatoryFeatures.end(),
[&](const string & feature) {
return features.count(feature);
2016-07-18 22:50:27 +00:00
});
}
Machine(decltype(storeUri) storeUri,
2016-07-18 22:50:27 +00:00
decltype(systemTypes) systemTypes,
decltype(sshKey) sshKey,
decltype(maxJobs) maxJobs,
decltype(speedFactor) speedFactor,
decltype(supportedFeatures) supportedFeatures,
decltype(mandatoryFeatures) mandatoryFeatures) :
2017-03-03 15:18:49 +00:00
supportedFeatures(supportedFeatures),
mandatoryFeatures(mandatoryFeatures),
storeUri(
// Backwards compatibility: if the URI is a hostname,
// prepend ssh://.
storeUri.find("://") != std::string::npos || hasPrefix(storeUri, "local") || hasPrefix(storeUri, "remote") || hasPrefix(storeUri, "auto")
? storeUri
: "ssh://" + storeUri),
2017-03-03 15:18:49 +00:00
systemTypes(systemTypes),
sshKey(sshKey),
maxJobs(maxJobs),
speedFactor(std::max(1U, speedFactor))
{}
};
2016-07-18 22:50:27 +00:00
2017-03-03 15:18:49 +00:00
static std::vector<Machine> readConf()
{
2016-07-18 22:50:27 +00:00
auto conf = getEnv("NIX_REMOTE_SYSTEMS", SYSCONFDIR "/nix/machines");
2017-03-03 15:18:49 +00:00
auto machines = std::vector<Machine>{};
auto lines = std::vector<string>{};
try {
lines = tokenizeString<std::vector<string>>(readFile(conf), "\n");
} catch (const SysError & e) {
if (e.errNo != ENOENT)
throw;
}
for (auto line : lines) {
chomp(line);
line.erase(std::find(line.begin(), line.end(), '#'), line.end());
if (line.empty()) {
continue;
}
auto tokens = tokenizeString<std::vector<string>>(line);
auto sz = tokens.size();
if (sz < 1)
2017-03-03 15:18:49 +00:00
throw FormatError("bad machines.conf file %1%", conf);
machines.emplace_back(tokens[0],
sz >= 2 ? tokenizeString<std::vector<string>>(tokens[1], ",") : std::vector<string>{settings.thisSystem},
sz >= 3 ? tokens[2] : "",
sz >= 4 ? std::stoull(tokens[3]) : 1LL,
sz >= 5 ? std::stoull(tokens[4]) : 1LL,
sz >= 6 ?
tokenizeString<std::set<string>>(tokens[5], ",") :
std::set<string>{},
sz >= 7 ?
tokenizeString<std::set<string>>(tokens[6], ",") :
std::set<string>{});
2016-07-18 22:50:27 +00:00
}
return machines;
}
std::string escapeUri(std::string uri)
{
std::replace(uri.begin(), uri.end(), '/', '_');
return uri;
}
2016-07-18 22:50:27 +00:00
static string currentLoad;
2017-03-03 15:18:49 +00:00
static AutoCloseFD openSlotLock(const Machine & m, unsigned long long slot)
{
return openLockFile(fmt("%s/%s-%d", currentLoad, escapeUri(m.storeUri), slot), true);
2016-07-18 22:50:27 +00:00
}
int main (int argc, char * * argv)
{
return handleExceptions(argv[0], [&]() {
initNix();
2017-03-03 15:18:49 +00:00
2016-07-18 22:50:27 +00:00
/* Ensure we don't get any SSH passphrase or host key popups. */
unsetenv("DISPLAY");
unsetenv("SSH_ASKPASS");
2016-07-18 22:50:27 +00:00
2017-05-01 12:43:14 +00:00
if (argc != 5)
2016-07-18 22:50:27 +00:00
throw UsageError("called without required arguments");
auto store = openStore();
auto localSystem = argv[1];
2017-05-01 12:43:14 +00:00
settings.maxSilentTime = std::stoll(argv[2]);
settings.buildTimeout = std::stoll(argv[3]);
verbosity = (Verbosity) std::stoll(argv[4]);
2016-07-18 22:50:27 +00:00
/* It would be more appropriate to use $XDG_RUNTIME_DIR, since
that gets cleared on reboot, but it wouldn't work on OS X. */
currentLoad = settings.nixStateDir + "/current-load";
2016-07-18 22:50:27 +00:00
std::shared_ptr<Store> sshStore;
AutoCloseFD bestSlotLock;
2017-03-03 15:18:49 +00:00
auto machines = readConf();
2017-05-01 12:43:14 +00:00
debug("got %d remote builders", machines.size());
2016-07-18 22:50:27 +00:00
string drvPath;
string storeUri;
2016-07-18 22:50:27 +00:00
for (string line; getline(cin, line);) {
auto tokens = tokenizeString<std::vector<string>>(line);
auto sz = tokens.size();
2017-03-03 15:18:49 +00:00
if (sz != 3 && sz != 4)
throw Error("invalid build hook line %1%", line);
2016-07-18 22:50:27 +00:00
auto amWilling = tokens[0] == "1";
auto neededSystem = tokens[1];
drvPath = tokens[2];
auto requiredFeatures = sz == 3 ?
std::set<string>{} :
tokenizeString<std::set<string>>(tokens[3], ",");
2016-07-18 22:50:27 +00:00
auto canBuildLocally = amWilling && (neededSystem == localSystem);
/* Error ignored here, will be caught later */
mkdir(currentLoad.c_str(), 0777);
while (true) {
bestSlotLock = -1;
AutoCloseFD lock = openLockFile(currentLoad + "/main-lock", true);
lockFile(lock.get(), ltWrite, true);
bool rightType = false;
2017-03-03 15:18:49 +00:00
Machine * bestMachine = nullptr;
2016-07-18 22:50:27 +00:00
unsigned long long bestLoad = 0;
for (auto & m : machines) {
debug("considering building on %s", m.storeUri);
2016-07-18 22:50:27 +00:00
if (m.enabled && std::find(m.systemTypes.begin(),
m.systemTypes.end(),
neededSystem) != m.systemTypes.end() &&
m.allSupported(requiredFeatures) &&
m.mandatoryMet(requiredFeatures)) {
rightType = true;
AutoCloseFD free;
unsigned long long load = 0;
for (unsigned long long slot = 0; slot < m.maxJobs; ++slot) {
2017-01-25 11:51:35 +00:00
auto slotLock = openSlotLock(m, slot);
2016-07-18 22:50:27 +00:00
if (lockFile(slotLock.get(), ltWrite, false)) {
if (!free) {
free = std::move(slotLock);
}
} else {
++load;
}
}
if (!free) {
continue;
}
bool best = false;
if (!bestSlotLock) {
best = true;
} else if (load / m.speedFactor < bestLoad / bestMachine->speedFactor) {
best = true;
} else if (load / m.speedFactor == bestLoad / bestMachine->speedFactor) {
if (m.speedFactor > bestMachine->speedFactor) {
best = true;
} else if (m.speedFactor == bestMachine->speedFactor) {
if (load < bestLoad) {
best = true;
}
}
}
if (best) {
bestLoad = load;
bestSlotLock = std::move(free);
bestMachine = &m;
}
}
}
if (!bestSlotLock) {
2017-03-03 15:18:49 +00:00
if (rightType && !canBuildLocally)
std::cerr << "# postpone\n";
else
std::cerr << "# decline\n";
2016-07-18 22:50:27 +00:00
break;
}
#if __APPLE__
futimes(bestSlotLock.get(), NULL);
#else
2016-07-18 22:50:27 +00:00
futimens(bestSlotLock.get(), NULL);
#endif
2016-07-18 22:50:27 +00:00
lock = -1;
try {
Store::Params storeParams{{"max-connections", "1"}};
if (bestMachine->sshKey != "")
storeParams["ssh-key"] = bestMachine->sshKey;
sshStore = openStore(bestMachine->storeUri, storeParams);
storeUri = bestMachine->storeUri;
2016-07-18 22:50:27 +00:00
} catch (std::exception & e) {
2017-03-03 15:18:49 +00:00
printError("unable to open SSH connection to %s: %s; trying other available machines...",
bestMachine->storeUri, e.what());
2016-07-18 22:50:27 +00:00
bestMachine->enabled = false;
continue;
}
2016-07-18 22:50:27 +00:00
goto connected;
}
}
2017-03-03 15:18:49 +00:00
2016-07-18 22:50:27 +00:00
connected:
2017-03-03 15:18:49 +00:00
std::cerr << "# accept\n";
2016-07-18 22:50:27 +00:00
string line;
2017-03-03 15:18:49 +00:00
if (!getline(cin, line))
2016-07-18 22:50:27 +00:00
throw Error("hook caller didn't send inputs");
auto inputs = tokenizeString<PathSet>(line);
2017-03-03 15:18:49 +00:00
if (!getline(cin, line))
2016-07-18 22:50:27 +00:00
throw Error("hook caller didn't send outputs");
auto outputs = tokenizeString<PathSet>(line);
AutoCloseFD uploadLock = openLockFile(currentLoad + "/" + escapeUri(storeUri) + ".upload-lock", true);
2017-03-03 15:18:49 +00:00
auto old = signal(SIGALRM, handleAlarm);
2016-07-18 22:50:27 +00:00
alarm(15 * 60);
2017-03-03 15:18:49 +00:00
if (!lockFile(uploadLock.get(), ltWrite, true))
printError("somebody is hogging the upload lock for %s, continuing...");
2016-07-18 22:50:27 +00:00
alarm(0);
signal(SIGALRM, old);
copyPaths(store, ref<Store>(sshStore), inputs, false, true);
2016-07-18 22:50:27 +00:00
uploadLock = -1;
BasicDerivation drv(readDerivation(drvPath));
drv.inputSrcs = inputs;
printError("building %s on %s", drvPath, storeUri);
sshStore->buildDerivation(drvPath, drv);
2016-07-18 22:50:27 +00:00
PathSet missing;
for (auto & path : outputs)
if (!store->isValidPath(path)) missing.insert(path);
if (!missing.empty()) {
setenv("NIX_HELD_LOCKS", concatStringsSep(" ", missing).c_str(), 1); /* FIXME: ugly */
copyPaths(ref<Store>(sshStore), store, missing, false, true);
2016-07-18 22:50:27 +00:00
}
2016-07-18 22:50:27 +00:00
return;
});
}