mgr: implement 'osd probably-ok-to-stop' command

This is an approximate check whether stopping the set of OSDs
concurrently will result in PGs becoming unavailable (by dropping
below min_size).

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

src/mgr/DaemonServer.cc

@@ -989,6 +989,78 @@ bool DaemonServer::handle_command(MCommand *m)
     ss << "osd." << osd << " is safe to destroy";
     cmdctx->reply(0, ss);
     return true;
+  } else if (prefix == "osd probably-ok-to-stop") {
+    vector<string> ids;
+    cmd_getval(g_ceph_context, cmdctx->cmdmap, "ids", ids);
+    set<int> osds;
+    int r;
+    cluster_state.with_osdmap([&](const OSDMap& osdmap) {
+	r = osdmap.parse_osd_id_list(ids, &osds, &ss);
+      });
+    if (r < 0) {
+      cmdctx->reply(r, ss);
+      return true;
+    }
+    map<pg_t,int> pg_delta;  // pgid -> net acting set size change
+    int dangerous_pgs = 0;
+    cluster_state.with_pgmap([&](const PGMap& pg_map) {
+	return cluster_state.with_osdmap([&](const OSDMap& osdmap) {
+	    if (pg_map.num_pg_unknown > 0) {
+	      ss << pg_map.num_pg_unknown << " pgs have unknown state; "
+		 << "cannot draw any conclusions";
+	      r = -EAGAIN;
+	      return;
+	    }
+	    for (auto osd : osds) {
+	      auto p = pg_map.pg_by_osd.find(osd);
+	      if (p != pg_map.pg_by_osd.end()) {
+		for (auto& pgid : p->second) {
+		  --pg_delta[pgid];
+		}
+	      }
+	    }
+	    for (auto& p : pg_delta) {
+	      auto q = pg_map.pg_stat.find(p.first);
+	      if (q == pg_map.pg_stat.end()) {
+		ss << "missing information about " << p.first << "; cannot draw"
+		   << " any conclusions";
+		r = -EAGAIN;
+		return;
+	      }
+	      if (!(q->second.state & PG_STATE_ACTIVE) ||
+		  (q->second.state & PG_STATE_DEGRADED)) {
+		// we don't currently have a good way to tell *how* degraded
+		// a degraded PG is, so we have to assume we cannot remove
+		// any more replicas/shards.
+		++dangerous_pgs;
+		return;
+	      }
+	      const pg_pool_t *pi = osdmap.get_pg_pool(p.first.pool());
+	      if (!pi) {
+		// pool deleted?
+	      } else {
+		if (q->second.acting.size() + p.second < pi->min_size) {
+		  ++dangerous_pgs;
+		}
+	      }
+	    }
+	  });
+      });
+    if (r) {
+      cmdctx->reply(r, ss);
+      return true;
+    }
+    if (dangerous_pgs) {
+      ss << dangerous_pgs << " are already degraded or might become unavailable";
+      cmdctx->reply(-EBUSY, ss);
+      return true;
+    }
+    ss << "OSD(s) " << osds << " are *probably* ok to stop without reducing "
+       << "availability, provided there are no other concurrent failures "
+       << "or interventions. " << pg_delta.size() << " PGs are likely to be "
+       << "degraded (but remain available) as a result.";
+    cmdctx->reply(0, ss);
+    return true;
   } else if (prefix == "pg force-recovery" ||
   	       prefix == "pg force-backfill" ||
   	       prefix == "pg cancel-force-recovery" ||

src/mgr/MgrCommands.h

@@ -110,6 +110,9 @@ COMMAND("osd test-reweight-by-pg " \
 COMMAND("osd safe-to-destroy name=id,type=CephOsdName",
 	"check whether osd.<id> can be safely destroyed without risking data loss",
 	"osd", "r", "cli,rest")
+COMMAND("osd probably-ok-to-stop name=ids,type=CephString,n=N",
+	"check whether osd(s) can be safely stopped without reducing availability",
+	"osd", "r", "cli,rest")
 
 COMMAND("osd scrub " \
 	"name=who,type=CephString", \