osd: add osd_fast_shutdown option (default true)

If we get a SIGINT or SIGTERM or are deleted from the OSDMap, do a fast
shutdown by exiting immediately.  This has a few important benefits:

 - We immediately stop responding (binding) to any sockets, which means
   other OSDs will immediately decide we are down (and dead!).  This
   minimizes IO interruption.
 - We avoid the complex "clean" shutdown process, which is historically a
   source of bugs.

In reality, the only purpose of the "clean" shutdown is to try to tear down
everything in memory so we can do memory leak checking with valgrind.  Set
this option to false for valgrind QA runs so we can still do that.

Not that with the new read leases in octopus, we rely on the default
behavior that a ECONNREFUSED is taken to mean that the OSD is fully dead,
so that we don't have to wait for any leases to time out.  This works in
sane environments with normal IP networks, but that behavior could
conceivably be a bad idea if there are some weird network shenanigans
going on.  If osd_fast_fail_on_connection_refused were disabled, then this
fast shutdown procedure might be *worse* than the clean shutdown because
we would have to wait for the heartbeat timeout.

Signed-off-by: Sage Weil <sage@redhat.com>

qa/suites/fs/verify/validater/valgrind.yaml

@@ -17,6 +17,8 @@ overrides:
         mds heartbeat grace: 60
       mon:
         mon osd crush smoke test: false
+      osd:
+        osd fast shutdown: false
     valgrind:
       mon: [--tool=memcheck, --leak-check=full, --show-reachable=yes]
       osd: [--tool=memcheck]

qa/suites/rados/singleton-flat/valgrind-leaks.yaml

@@ -23,6 +23,8 @@ overrides:
         osd max object namespace len: 64
       mon:
         mon osd crush smoke test: false
+      osd:
+        osd fast shutdown: false
     valgrind:
       mon: [--tool=memcheck, --leak-check=full, --show-reachable=yes]
       osd: [--tool=memcheck]

qa/suites/rados/verify/validater/valgrind.yaml

@@ -13,6 +13,8 @@ overrides:
         debug refs: 5
       mon:
         mon osd crush smoke test: false
+      osd:
+        osd fast shutdown: false
     log-whitelist:
       - overall HEALTH_
 # valgrind is slow.. we might get PGs stuck peering etc

qa/suites/rgw/multisite/valgrind.yaml

@@ -11,6 +11,8 @@ overrides:
         osd heartbeat grace: 40
       mon:
         mon osd crush smoke test: false
+      osd:
+        osd fast shutdown: false
     valgrind:
       mon: [--tool=memcheck, --leak-check=full, --show-reachable=yes]
       osd: [--tool=memcheck]

qa/suites/rgw/verify/validater/valgrind.yaml

@@ -12,6 +12,8 @@ overrides:
         osd heartbeat grace: 40
       mon:
         mon osd crush smoke test: false
+      osd:
+        osd fast shutdown: false
     valgrind:
       mon: [--tool=memcheck, --leak-check=full, --show-reachable=yes]
       osd: [--tool=memcheck]

src/common/legacy_config_opts.h

@@ -764,6 +764,7 @@ OPTION(osd_op_history_slow_op_size, OPT_U32)           // Max number of slow ops
 OPTION(osd_op_history_slow_op_threshold, OPT_DOUBLE) // track the op if over this threshold
 OPTION(osd_target_transaction_size, OPT_INT)     // to adjust various transactions that batch smaller items
 OPTION(osd_failsafe_full_ratio, OPT_FLOAT) // what % full makes an OSD "full" (failsafe)
+OPTION(osd_fast_shutdown, OPT_BOOL)
 OPTION(osd_fast_fail_on_connection_refused, OPT_BOOL) // immediately mark OSDs as down once they refuse to accept connections
 
 OPTION(osd_pg_object_context_cache_count, OPT_INT)

src/common/options.cc

@@ -3498,6 +3498,11 @@ std::vector<Option> get_global_options() {
     .set_default(.97)
     .set_description(""),
 
+    Option("osd_fast_shutdown", Option::TYPE_BOOL, Option::LEVEL_ADVANCED)
+    .set_default(true)
+    .set_description("Fast, immediate shutdown")
+    .set_long_description("Setting this to false makes the OSD do a slower teardown of all state when it receives a SIGINT or SIGTERM or when shutting down for any other reason.  That slow shutdown is primarilyy useful for doing memory leak checking with valgrind."),
+
     Option("osd_fast_fail_on_connection_refused", Option::TYPE_BOOL, Option::LEVEL_ADVANCED)
     .set_default(true)
     .set_description(""),

src/osd/OSD.cc

@@ -4016,6 +4016,12 @@ void OSD::create_recoverystate_perf()
 
 int OSD::shutdown()
 {
+  if (cct->_conf->osd_fast_shutdown) {
+    derr << "*** Immediate shutdown (osd_fast_shutdown=true) ***" << dendl;
+    cct->_log->flush();
+    _exit(0);
+  }
+
   if (!service.prepare_to_stop())
     return 0; // already shutting down
   osd_lock.lock();

src/vstart.sh

@@ -710,6 +710,7 @@ $DAEMONOPTS
         osd class dir = $OBJCLASS_PATH
         osd class load list = *
         osd class default list = *
+        osd fast shutdown = false
 
         filestore wbthrottle xfs ios start flusher = 10
         filestore wbthrottle xfs ios hard limit = 20