mgr/DaemonServer: osd ok-to-stop: return json when there are unknown PGs

In 791952cc01201010f298033003ba52374cc0159f we switched to return JSON
both on success and fail to describe which PGs are affected or are blocking
the ability to stop/restart OSDs.  Do the same for the case where
some PG states are unknown (i.e., just after a mgr restart) so that
the cephadm upgrade process can unconditionally expect a JSON result.

Signed-off-by: Sage Weil <sage@newdream.net>

diff --git a/src/mgr/DaemonServer.cc b/src/mgr/DaemonServer.cc
index 88f325f10e4285aa6c3efc58d0116012723f77a7..b70ea066680e8365e758e5f3e401cd8c72aa40e0 100644 (file)
--- a/src/mgr/DaemonServer.cc
+++ b/src/mgr/DaemonServer.cc
@@ -881,6 +881,10 @@ void DaemonServer::_check_offlines_pgs(
   for (const auto& q : pgmap.pg_stat) {
     set<int32_t> pg_acting;  // net acting sets (with no missing if degraded)
     bool found = false;
+    if (q.second.state == 0) {
+      report->unknown.insert(q.first);
+      continue;
+    }
     if (q.second.state & PG_STATE_DEGRADED) {
       for (auto& anm : q.second.avail_no_missing) {
        if (osds.count(anm.osd)) {
@@ -931,7 +935,9 @@ void DaemonServer::_check_offlines_pgs(
     }
   }
   dout(20) << osds << " -> " << report->ok.size() << " ok, "
-          << report->not_ok.size() << " not ok" << dendl;
+          << report->not_ok.size() << " not ok, "
+          << report->unknown.size() << " unknown"
+          << dendl;
 }
 
 void DaemonServer::_maximize_ok_to_stop_set(
@@ -1737,26 +1743,21 @@ bool DaemonServer::_handle_command(
     }
     offline_pg_report out_report;
     cluster_state.with_osdmap_and_pgmap([&](const OSDMap& osdmap, const PGMap& pg_map) {
-       if (pg_map.num_pg_unknown > 0) {
-         ss << pg_map.num_pg_unknown << " pgs have unknown state; "
-            << "cannot draw any conclusions";
-         r = -EAGAIN;
-         return;
-       }
        _maximize_ok_to_stop_set(
          osds, max, osdmap, pg_map,
          &out_report);
       });
-    if (r < 0) {
-      cmdctx->reply(r, ss);
-      return true;
-    }
     if (!f) {
       f.reset(Formatter::create("json"));
     }
     f->dump_object("ok_to_stop", out_report);
     f->flush(cmdctx->odata);
     cmdctx->odata.append("\n");
+    if (!out_report.unknown.empty()) {
+      ss << out_report.unknown.size() << " pgs have unknown state; "
+        << "cannot draw any conclusions";
+      cmdctx->reply(-EAGAIN, ss);
+    }
     if (!out_report.ok_to_stop()) {
       ss << "unsafe to stop osd(s)";
       cmdctx->reply(-EBUSY, ss);

diff --git a/src/mgr/DaemonServer.h b/src/mgr/DaemonServer.h
index 9119c6743f86a5f0c897c2ed7760e8e51f584785..3adcf6ccaf93b8d50af5cd15202b115492e4340a 100644 (file)
--- a/src/mgr/DaemonServer.h
+++ b/src/mgr/DaemonServer.h
@@ -48,12 +48,12 @@ struct MDSPerfMetricQuery;
 
 struct offline_pg_report {
   set<int> osds;
-  set<pg_t> ok, not_ok;
+  set<pg_t> ok, not_ok, unknown;
   set<pg_t> ok_become_degraded, ok_become_more_degraded;             // ok
   set<pg_t> bad_no_pool, bad_already_inactive, bad_become_inactive;  // not ok
 
   bool ok_to_stop() const {
-    return not_ok.empty();
+    return not_ok.empty() && unknown.empty();
   }
 
   void dump(Formatter *f) const {
@@ -66,6 +66,15 @@ struct offline_pg_report {
     f->dump_unsigned("num_ok_pgs", ok.size());
     f->dump_unsigned("num_not_ok_pgs", not_ok.size());
 
+    // ambiguous
+    if (!unknown.empty()) {
+      f->open_array_section("unknown_pgs");
+      for (auto pg : unknown) {
+       f->dump_stream("pg") << pg;
+      }
+      f->close_section();
+    }
+
     // bad news
     if (!bad_no_pool.empty()) {
       f->open_array_section("bad_no_pool_pgs");