mds: skip journaling blocklisted clients when in `replay` state

When a standby MDS is transitioning to active, it passes through
`replay` state. When the MDS is in this state, there are no journal
segments available for recording journal updates. If the MDS receives
an OSDMap update in this state, journaling blocklisted clients causes
a crash since no journal segments are available. This is a bit hard
to reproduce as it requires correct timing of an OSDMap update along
with various other factors.

Note that, when the MDS reaches `reconnect` state, it will journal
the blocklisted clients anyway.

This partially fixes tracker: https://tracker.ceph.com/issues/51589
which mentions a similar crash but in `reconnect` state. However,
that crash was seen in nautilus.

A couple of minor changes include removing hardcoded function names
and carving out reusable parts into a separate function.

Partially-fixes: https://tracker.ceph.com/issues/51589
Signed-off-by: Venky Shankar <vshankar@redhat.com>

diff --git a/src/mds/MDSRank.cc b/src/mds/MDSRank.cc
index 190a53706455d6075b24b57f2187c1d4bf9aed8a..f5f0585060dabf4768bab5b89ff791325efee25e 100644 (file)
--- a/src/mds/MDSRank.cc
+++ b/src/mds/MDSRank.cc
@@ -1930,6 +1930,17 @@ void MDSRank::resolve_done()
   snapclient->sync(new C_MDSInternalNoop);
 }
 
+void MDSRank::apply_blocklist(const std::set<entity_addr_t> &addrs, epoch_t epoch) {
+  auto victims = server->apply_blocklist(addrs);
+  dout(4) << __func__ << ": killed " << victims << " blocklisted sessions ("
+          << addrs.size() << " blocklist entries, "
+          << sessionmap.get_sessions().size() << ")" << dendl;
+  if (victims) {
+    set_osd_epoch_barrier(epoch);
+  }
+}
+
+
 void MDSRank::reconnect_start()
 {
   dout(1) << "reconnect_start" << dendl;
@@ -1947,13 +1958,8 @@ void MDSRank::reconnect_start()
       o.get_blocklist(&blocklist);
       epoch = o.get_epoch();
   });
-  auto killed = server->apply_blocklist(blocklist);
-  dout(4) << "reconnect_start: killed " << killed << " blocklisted sessions ("
-          << blocklist.size() << " blocklist entries, "
-          << sessionmap.get_sessions().size() << ")" << dendl;
-  if (killed) {
-    set_osd_epoch_barrier(epoch);
-  }
+
+  apply_blocklist(blocklist, epoch);
 
   server->reconnect_clients(new C_MDS_VoidFn(this, &MDSRank::reconnect_done));
   finish_contexts(g_ceph_context, waiting_for_reconnect);
@@ -3455,17 +3461,17 @@ void MDSRankDispatcher::handle_osd_map()
 
   purge_queue.update_op_limit(*mdsmap);
 
-  std::set<entity_addr_t> newly_blocklisted;
-  objecter->consume_blocklist_events(&newly_blocklisted);
-  auto epoch = objecter->with_osdmap([](const OSDMap &o){return o.get_epoch();});
-  dout(4) << "handle_osd_map epoch " << epoch << ", "
-          << newly_blocklisted.size() << " new blocklist entries" << dendl;
-  auto victims = server->apply_blocklist(newly_blocklisted);
-  if (victims) {
-    set_osd_epoch_barrier(epoch);
+  // it's ok if replay state is reached via standby-replay, the
+  // reconnect state will journal blocklisted clients (journal
+  // is opened for writing in `replay_done` before moving to
+  // up:resolve).
+  if (!is_replay()) {
+    std::set<entity_addr_t> newly_blocklisted;
+    objecter->consume_blocklist_events(&newly_blocklisted);
+    auto epoch = objecter->with_osdmap([](const OSDMap &o){return o.get_epoch();});
+    apply_blocklist(newly_blocklisted, epoch);
   }
 
-
   // By default the objecter only requests OSDMap updates on use,
   // we would like to always receive the latest maps in order to
   // apply policy based on the FULL flag.

diff --git a/src/mds/MDSRank.h b/src/mds/MDSRank.h
index 6ba2001932867131206afad769053de531254266..21b5ac7924b62b1658def11bcbddef5852dd5429 100644 (file)
--- a/src/mds/MDSRank.h
+++ b/src/mds/MDSRank.h
@@ -546,6 +546,10 @@ class MDSRank {
 
     Context *create_async_exec_context(C_ExecAndReply *ctx);
 
+    // blocklist the provided addrs and set OSD epoch barrier
+    // with the provided epoch.
+    void apply_blocklist(const std::set<entity_addr_t> &addrs, epoch_t epoch);
+
     // Incarnation as seen in MDSMap at the point where a rank is
     // assigned.
     int incarnation = 0;