Make leader to delete the raft repl dev in UT.

conanfile.py

@@ -9,7 +9,7 @@
 
 class HomestoreConan(ConanFile):
     name = "homestore"
-    version = "6.4.45"
+    version = "6.4.46"
 
     homepage = "https://github.com/eBay/Homestore"
     description = "HomeStore Storage Engine"

src/lib/replication/repl_dev/raft_repl_dev.cpp

@@ -230,20 +230,20 @@ void RaftReplDev::push_data_to_all_followers(repl_req_ptr_t rreq, sisl::sg_list
            flatbuffers::FlatBufferToString(builder.GetBufferPointer() + sizeof(flatbuffers::uoffset_t),
                                            PushDataRequestTypeTable()));*/
 
-    RD_LOGD("Data Channel: Pushing data to all followers: rreq=[{}]", rreq->to_compact_string());
+    RD_LOGD("Data Channel: Pushing data to all followers: rreq=[{}]", rreq->to_string());
 
     group_msg_service()
         ->data_service_request_unidirectional(nuraft_mesg::role_regex::ALL, PUSH_DATA, rreq->m_pkts)
         .via(&folly::InlineExecutor::instance())
         .thenValue([this, rreq = std::move(rreq)](auto e) {
             if (e.hasError()) {
-                RD_LOGE("Data Channel: Error in pushing data to all followers: rreq=[{}] error={}",
-                        rreq->to_compact_string(), e.error());
+                RD_LOGE("Data Channel: Error in pushing data to all followers: rreq=[{}] error={}", rreq->to_string(),
+                        e.error());
                 handle_error(rreq, RaftReplService::to_repl_error(e.error()));
                 return;
             }
             // Release the buffer which holds the packets
-            RD_LOGD("Data Channel: Data push completed for rreq=[{}]", rreq->to_compact_string());
+            RD_LOGD("Data Channel: Data push completed for rreq=[{}]", rreq->to_string());
             rreq->release_fb_builder();
             rreq->m_pkts.clear();
         });

src/lib/replication/service/raft_repl_service.cpp

@@ -14,6 +14,7 @@
  *********************************************************************************/
 #include <sisl/logging/logging.h>
 #include <iomgr/io_environment.hpp>
+#include <iomgr/iomgr_flip.hpp>
 #include <chrono>
 
 #include <boost/uuid/string_generator.hpp>
@@ -118,6 +119,7 @@ void RaftReplService::start() {
     // We need to first load the repl_dev with its config and then attach the raft config to that repl dev.
     for (auto const& [buf, mblk] : m_config_sb_bufs) {
         auto rdev = raft_group_config_found(buf, voidptr_cast(mblk));
+        if (!rdev) { continue; }
         rdev->on_restart();
     }
     m_config_sb_bufs.clear();
@@ -342,6 +344,9 @@ void RaftReplService::start_reaper_thread() {
                 HS_DYNAMIC_CONFIG(generic.repl_dev_cleanup_interval_sec) * 1000 * 1000 * 1000, true /* recurring */,
                 nullptr, [this](void*) {
                     gc_repl_reqs();
+#ifdef _PRERELEASE
+                    if (iomgr_flip::instance()->test_flip("disable_periodic_gc_repl_dev")) { return; }
+#endif
                     gc_repl_devs();
                 });
 

src/tests/test_raft_repl_dev.cpp

@@ -391,6 +391,11 @@ class RaftReplDevTest : public testing::Test {
             });
         }
 
+        if (g_helper->replica_num() == 0) {
+            // Leader sleeps here because follower-1 needs some time to find the leader after restart.
+            std::this_thread::sleep_for(std::chrono::seconds(5));
+        }
+
         for (auto const& db : dbs_) {
             if (db->is_zombie()) { continue; }
             auto repl_dev = std::dynamic_pointer_cast< RaftReplDev >(db->repl_dev());
@@ -767,6 +772,7 @@ TEST_F(RaftReplDevTest, Snapshot_and_Compact) {
 
 TEST_F(RaftReplDevTest, RemoveReplDev) {
     LOGINFO("Homestore replica={} setup completed", g_helper->replica_num());
+    g_helper->set_basic_flip("disable_periodic_gc_repl_dev");
 
     // Step 1: Create 2 more repldevs
     LOGINFO("Create 2 more ReplDevs");
@@ -806,11 +812,6 @@ TEST_F(RaftReplDevTest, RemoveReplDev) {
         LOGINFO("Remove last replica={} num_db={}", g_helper->replica_num(), dbs_.size());
     }
 
-    if (g_helper->replica_num() == 0) {
-        // Leader sleeps here because follower-1 needs some time to find the leader after restart.
-        std::this_thread::sleep_for(std::chrono::seconds(20));
-    }
-
     // TODO: Once generic crash flip/test_infra is available, use flip to crash during removal and restart them to
     // see if records are being removed
     g_helper->sync_for_cleanup_start();
@@ -944,7 +945,7 @@ int main(int argc, char* argv[]) {
     //
     HS_SETTINGS_FACTORY().modifiable_settings([](auto& s) {
         s.consensus.leadership_expiry_ms = -1; // -1 means never expires;
-        s.generic.repl_dev_cleanup_interval_sec = 1;
+        s.generic.repl_dev_cleanup_interval_sec = 0;
 
         // Disable implicit flush and timer.
         s.logstore.flush_threshold_size = 0;