Merge pull request ceph#60067 from VallariAg/wip-nvmeof-healthcheck

mon: add nvmeof healthchecks

doc/rados/operations/health-checks.rst

@@ -1640,6 +1640,25 @@ We encourage you to fix this by making the weights even on both dividing buckets
 This can be done by making sure the combined weight of the OSDs on each dividing
 bucket are the same.
 
+NVMeoF Gateway
+--------------
+
+NVMEOF_SINGLE_GATEWAY
+_____________________
+
+One of the gateway group has only one gateway. This is not ideal because it makes
+high availability (HA) impossible with a single gatway in a group. This can lead to 
+problems with failover and failback operations for the NVMeoF gateway.
+
+It's recommended to have multiple NVMeoF gateways in a group.
+
+NVMEOF_GATEWAY_DOWN
+___________________
+
+Some of the gateways are in the GW_UNAVAILABLE state. If a NVMeoF daemon has crashed, 
+the daemon log file (found at ``/var/log/ceph/``) may contain troubleshooting information.
+
+
 Miscellaneous
 -------------
 

qa/suites/nvmeof/basic/clusters/4-gateways-2-initiator.yaml

@@ -28,3 +28,5 @@ overrides:
       mon:
         # cephadm can take up to 5 minutes to bring up remaining mons
         mon down mkfs grace: 300
+    log-ignorelist:
+      - NVMEOF_SINGLE_GATEWAY

qa/suites/nvmeof/thrash/thrashers/nvmeof_mon_thrash.yaml

@@ -8,6 +8,9 @@ overrides:
       - out of quorum
       # nvmeof daemon thrashing
       - CEPHADM_FAILED_DAEMON
+      - NVMEOF_SINGLE_GATEWAY
+      - NVMEOF_GATEWAY_DOWN
+      - are in unavailable state
       - is in error state
       - failed cephadm daemon
 

qa/suites/nvmeof/thrash/thrashers/nvmeof_thrash.yaml

@@ -3,6 +3,9 @@ overrides:
     log-ignorelist:  
       # nvmeof daemon thrashing
       - CEPHADM_FAILED_DAEMON
+      - NVMEOF_SINGLE_GATEWAY
+      - NVMEOF_GATEWAY_DOWN
+      - are in unavailable state
       - is in error state
       - failed cephadm daemon
 

src/mon/NVMeofGwMap.cc

@@ -16,7 +16,9 @@
 #include "NVMeofGwMon.h"
 #include "NVMeofGwMap.h"
 #include "OSDMonitor.h"
+#include "mon/health_check.h"
 
+using std::list;
 using std::map;
 using std::make_pair;
 using std::ostream;
@@ -893,6 +895,47 @@ struct CMonRequestProposal : public Context {
   }
 };
 
+void NVMeofGwMap::get_health_checks(health_check_map_t *checks) const 
+{
+  list<string> singleGatewayDetail;
+  list<string> gatewayDownDetail;
+  for (const auto& created_map_pair: created_gws) {
+    const auto& group_key = created_map_pair.first;
+    auto& group = group_key.second;
+    const NvmeGwMonStates& gw_created_map = created_map_pair.second;
+    if ( gw_created_map.size() == 1) {
+      ostringstream ss;
+      ss << "NVMeoF Gateway Group '" << group << "' has 1 gateway." ;
+      singleGatewayDetail.push_back(ss.str());
+    }
+    for (const auto& gw_created_pair: gw_created_map) {
+      const auto& gw_id = gw_created_pair.first;
+      const auto& gw_created  = gw_created_pair.second;
+      if (gw_created.availability == gw_availability_t::GW_UNAVAILABLE) {
+        ostringstream ss;
+        ss << "NVMeoF Gateway '" << gw_id << "' is unavailable." ;
+        gatewayDownDetail.push_back(ss.str());
+      }
+    }
+  }
+  if (!singleGatewayDetail.empty()) {
+    ostringstream ss;
+    ss << singleGatewayDetail.size() << " group(s) have only 1 nvmeof gateway"
+      << "; HA is not possible with single gateway.";
+    auto& d = checks->add("NVMEOF_SINGLE_GATEWAY", HEALTH_WARN,
+        ss.str(), singleGatewayDetail.size());
+    d.detail.swap(singleGatewayDetail);
+  }
+  if (!gatewayDownDetail.empty()) {
+    ostringstream ss;
+    ss << gatewayDownDetail.size() << " gateway(s) are in unavailable state"
+      << "; gateway might be down, try to redeploy.";
+    auto& d = checks->add("NVMEOF_GATEWAY_DOWN", HEALTH_WARN,
+        ss.str(), gatewayDownDetail.size());
+    d.detail.swap(gatewayDownDetail);
+  }
+}
+
 int NVMeofGwMap::blocklist_gw(
   const NvmeGwId &gw_id, const NvmeGroupKey& group_key,
   NvmeAnaGrpId grpid, epoch_t &epoch, bool failover)

src/mon/NVMeofGwMap.h

@@ -27,6 +27,9 @@
 #include "NVMeofGwTypes.h"
 
 using ceph::coarse_mono_clock;
+
+class health_check_map_t;
+
 class Monitor;
 /*-------------------*/
 class NVMeofGwMap
@@ -140,6 +143,8 @@ class NVMeofGwMap
     decode(fsm_timers, bl);
     DECODE_FINISH(bl);
   }
+
+  void get_health_checks(health_check_map_t *checks) const;
 };
 
 #include "NVMeofGwSerialize.h"

src/mon/NVMeofGwMon.cc

@@ -181,6 +181,11 @@ void NVMeofGwMon::encode_pending(MonitorDBStore::TransactionRef t)
 	   << HAVE_FEATURE(mon.get_quorum_con_features(), NVMEOFHA) << dendl;
   put_version(t, pending_map.epoch, bl);
   put_last_committed(t, pending_map.epoch);
+
+  //health
+  health_check_map_t checks;
+  pending_map.get_health_checks(&checks);
+  encode_health(checks, t);
 }
 
 void NVMeofGwMon::update_from_paxos(bool *need_bootstrap)
@@ -193,6 +198,7 @@ void NVMeofGwMon::update_from_paxos(bool *need_bootstrap)
     bufferlist bl;
     int err = get_version(version, bl);
     ceph_assert(err == 0);
+    load_health();
 
     auto p = bl.cbegin();
     map.decode(p);