Configure keep-alive for server sockets.

Minor fixes to docker build.
Provide a workaround for #35 that should suffice for the initial release.

helio

@@ -1 +1 @@
-Subproject commit 3b83acc66bf7bc2308435039b0aba565626d933a
+Subproject commit 0420a22a085f09b01ff9a0849f70779492988039

src/core/expire_period.h

@@ -13,7 +13,7 @@ class ExpirePeriod {
   static constexpr size_t kMaxGenId = 15;
 
   ExpirePeriod() : val_(0), gen_(0), precision_(0) {
-    static_assert(sizeof(ExpirePeriod) == 4);
+    static_assert(sizeof(ExpirePeriod) == 8);  // TODO
   }
 
   explicit ExpirePeriod(uint64_t ms, unsigned gen = 0) : ExpirePeriod() {
@@ -38,13 +38,13 @@ class ExpirePeriod {
   bool is_second_precision() { return precision_ == 1;}
 
  private:
-  uint32_t val_ : 27;
-  uint32_t gen_ : 4;
-  uint32_t precision_ : 1;  // 0 - ms, 1 - sec.
+  uint64_t val_ : 59;
+  uint64_t gen_ : 4;
+  uint64_t precision_ : 1;  // 0 - ms, 1 - sec.
 };
 
 inline void ExpirePeriod::Set(uint64_t ms) {
-  constexpr uint64_t kBarrier = (1ULL << 27);
+  constexpr uint64_t kBarrier = (1ULL << 48);
 
   if (ms < kBarrier) {
     val_ = ms;

src/facade/dragonfly_listener.cc

@@ -42,6 +42,7 @@ namespace facade {
 using namespace util;
 using namespace std;
 
+namespace {
 // To connect: openssl s_client  -cipher "ADH:@SECLEVEL=0" -state -crlf  -connect 127.0.0.1:6380
 static SSL_CTX* CreateSslCntx() {
   SSL_CTX* ctx = SSL_CTX_new(TLS_server_method());
@@ -86,6 +87,35 @@ static SSL_CTX* CreateSslCntx() {
   return ctx;
 }
 
+bool ConfigureKeepAlive(int fd, unsigned interval_sec) {
+  DCHECK_GT(interval_sec, 3u);
+
+  int val = 1;
+  if (setsockopt(fd, SOL_SOCKET, SO_KEEPALIVE, &val, sizeof(val)) < 0)
+    return false;
+
+  val = interval_sec;
+  if (setsockopt(fd, IPPROTO_TCP, TCP_KEEPIDLE, &val, sizeof(val)) < 0)
+    return false;
+
+  /* Send next probes after the specified interval. Note that we set the
+    * delay as interval / 3, as we send three probes before detecting
+    * an error (see the next setsockopt call). */
+  val = interval_sec / 3;
+  if (setsockopt(fd, IPPROTO_TCP, TCP_KEEPINTVL, &val, sizeof(val)) < 0)
+    return false;
+
+  /* Consider the socket in error state after three we send three ACK
+    * probes without getting a reply. */
+  val = 3;
+  if (setsockopt(fd, IPPROTO_TCP, TCP_KEEPCNT, &val, sizeof(val)) < 0)
+    return false;
+
+  return true;
+}
+
+}  // namespace
+
 Listener::Listener(Protocol protocol, ServiceInterface* e) : service_(e), protocol_(protocol) {
   if (FLAGS_tls) {
     OPENSSL_init_ssl(OPENSSL_INIT_SSL_DEFAULT, NULL);
@@ -104,6 +134,20 @@ util::Connection* Listener::NewConnection(ProactorBase* proactor) {
   return new Connection{protocol_, http_base_.get(), ctx_, service_};
 }
 
+error_code Listener::ConfigureServerSocket(int fd) {
+  int val = 1;
+  constexpr int kInterval = 300;  // 300 seconds is ok to start checking for liveness.
+
+  if (setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &val, sizeof(val)) < 0) {
+    LOG(WARNING) << "Could not set reuse addr on socket " << detail::SafeErrorMessage(errno);
+  }
+  bool success = ConfigureKeepAlive(fd, kInterval);
+
+  LOG_IF(WARNING, !success) << "Could not configure keep alive " << detail::SafeErrorMessage(errno);
+
+  return error_code{};
+}
+
 void Listener::PreShutdown() {
 }
 

src/facade/dragonfly_listener.h

@@ -19,13 +19,15 @@ class Listener : public util::ListenerInterface {
   Listener(Protocol protocol, ServiceInterface*);
   ~Listener();
 
+  std::error_code ConfigureServerSocket(int fd) final;
+
  private:
   util::Connection* NewConnection(util::ProactorBase* proactor) final;
   util::ProactorBase* PickConnectionProactor(util::LinuxSocketBase* sock) final;
 
-  void PreShutdown();
+  void PreShutdown() final;
 
-  void PostShutdown();
+  void PostShutdown() final;
 
   std::unique_ptr<util::HttpListenerBase> http_base_;
 

src/server/db_slice.cc

@@ -32,7 +32,8 @@ constexpr auto kExpireSegmentSize = ExpireTable::kSegBytes;
 static_assert(kPrimeSegmentSize == 32720);
 
 // 20480 is the next goodsize so we are loosing ~300 bytes or 1.5%.
-static_assert(kExpireSegmentSize == 20168);
+// 24576
+static_assert(kExpireSegmentSize == 23528);
 
 class PrimeEvictionPolicy {
  public:

tools/docker/Dockerfile.ubuntu-prod

@@ -20,12 +20,15 @@ RUN curl -O https://raw.githubusercontent.com/ncopa/su-exec/master/su-exec.c &&
 # Now prod image
 FROM ubuntu:20.04
 
+# ARG in fact change the env vars during the build process
+# ENV persist the env vars for the built image as well.
 ARG QEMU_CPU
 ARG ORG_NAME=dragonflydb
+ARG DEBIAN_FRONTEND=noninteractive
 
 LABEL org.opencontainers.image.title Dragonfly
 LABEL org.opencontainers.image.source https://github.com/${ORG_NAME}/dragonfly
-ENV DEBIAN_FRONTEND=noninteractive
+
 
 RUN groupadd -r -g 999 dfly && useradd -r -g dfly -u 999 dfly
 RUN apt update && apt install -y libunwind8 libssl1.1 && \
@@ -40,5 +43,7 @@ COPY --from=builder /build/build-opt/su-exec /usr/local/bin/
 
 ENTRYPOINT ["entrypoint.sh"]
 
-EXPOSE 6380
+# For inter-container communication.
+EXPOSE 6379
+
 CMD ["dragonfly", "--logtostderr"]