This is an automated email from the ASF dual-hosted git repository.
gavinchou pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/doris.git
The following commit(s) were added to refs/heads/master by this push:
new 22b322c6edf [fix](cloud) Preserve clone source mappings for deleted
instances (#68343)
22b322c6edf is described below
commit 22b322c6edf023c43c69424ceb1b6cf8f7b509ba
Author: Yixuan Wang <[email protected]>
AuthorDate: Mon Sep 28 18:01:04 2026 +0800
[fix](cloud) Preserve clone source mappings for deleted instances (#68343)
Update source snapshot mappings before handling DELETED instances so a
deleted intermediate clone can still be traversed by descendants.
Continue clearing node routing and multi-version runtime state for
deleted instances, while keeping valid clone lineage available for
rollback recovery.
Add resource-manager lifecycle coverage and a multi-level clone-chain
regression test.
---
cloud/src/resource-manager/resource_manager.cpp | 9 ----
cloud/test/clone_chain_reader_test.cpp | 62 +++++++++++++++++++++++++
cloud/test/resource_test.cpp | 10 ++--
3 files changed, 67 insertions(+), 14 deletions(-)
diff --git a/cloud/src/resource-manager/resource_manager.cpp
b/cloud/src/resource-manager/resource_manager.cpp
index 3c5f52a7bad..fafd4ae09e8 100644
--- a/cloud/src/resource-manager/resource_manager.cpp
+++ b/cloud/src/resource-manager/resource_manager.cpp
@@ -109,9 +109,6 @@ int ResourceManager::init() {
std::unique_lock l(mtx_);
for (auto& [inst_id, inst] : instances) {
- if (inst.status() == InstanceInfoPB::DELETED) {
- continue;
- }
for (auto& c : inst.clusters()) {
add_cluster_to_index_no_lock(inst_id, c);
}
@@ -1442,12 +1439,6 @@ void ResourceManager::refresh_instance(const
std::string& instance_id,
}
}
- if (instance.status() == InstanceInfoPB::DELETED) {
- instance_multi_version_status_.erase(instance_id);
- instance_source_snapshot_info_.erase(instance_id);
- return;
- }
-
// If successor_instance_id is set, it means this instance has a successor
instance,
// so we do not need to add its clusters to the index again.
if (!instance.has_successor_instance_id()) {
diff --git a/cloud/test/clone_chain_reader_test.cpp
b/cloud/test/clone_chain_reader_test.cpp
index a94e531b57e..089747faa36 100644
--- a/cloud/test/clone_chain_reader_test.cpp
+++ b/cloud/test/clone_chain_reader_test.cpp
@@ -2232,6 +2232,68 @@ TEST_F(CloneChainReaderTest, GetTabletMeta) {
}
}
+TEST_F(CloneChainReaderTest, GetTabletMetaThroughDeletedIntermediateClone) {
+ ResourceManager resource_mgr(txn_kv_);
+
+ InstanceInfoPB instance_b;
+ instance_b.set_instance_id(instance_ids_[1]);
+ instance_b.set_status(InstanceInfoPB::NORMAL);
+ instance_b.set_source_instance_id(instance_ids_[0]);
+
instance_b.set_source_snapshot_id(Versionstamp(snapshot_versions_[0]).to_string());
+ resource_mgr.refresh_instance(instance_ids_[1], instance_b);
+
+ InstanceInfoPB instance_c;
+ instance_c.set_instance_id(instance_ids_[2]);
+ instance_c.set_status(InstanceInfoPB::NORMAL);
+ instance_c.set_source_instance_id(instance_ids_[1]);
+
instance_c.set_source_snapshot_id(Versionstamp(snapshot_versions_[1]).to_string());
+ resource_mgr.refresh_instance(instance_ids_[2], instance_c);
+
+ constexpr int64_t tablet_id = 16002;
+ {
+ std::unique_ptr<Transaction> txn;
+ ASSERT_EQ(txn_kv_->create_txn(&txn), TxnErrorCode::TXN_OK);
+ std::string tablet_meta_key =
versioned::meta_tablet_key({instance_ids_[0], tablet_id});
+ doris::TabletMetaCloudPB tablet_meta;
+ tablet_meta.set_tablet_id(tablet_id);
+ tablet_meta.set_table_id(1001);
+ ASSERT_TRUE(versioned::document_put(txn.get(), tablet_meta_key,
std::move(tablet_meta)));
+ ASSERT_EQ(txn->commit(), TxnErrorCode::TXN_OK);
+ }
+
+ // Single-level clone lookup remains unchanged.
+ {
+ CloneChainReader reader(instance_ids_[1],
Versionstamp(snapshot_versions_[1]),
+ txn_kv_.get(), &resource_mgr);
+ doris::TabletMetaCloudPB tablet_meta;
+ Versionstamp versionstamp;
+ ASSERT_EQ(reader.get_tablet_meta(tablet_id, &tablet_meta,
&versionstamp),
+ TxnErrorCode::TXN_OK);
+ EXPECT_EQ(tablet_meta.table_id(), 1001);
+ }
+
+ instance_b.set_status(InstanceInfoPB::DELETED);
+ resource_mgr.refresh_instance(instance_ids_[1], instance_b);
+
+ // B's tombstone still carries B -> A lineage, so C can continue to A.
+ CloneChainReader reader(instance_ids_[2],
Versionstamp(snapshot_versions_[2]), txn_kv_.get(),
+ &resource_mgr);
+ doris::TabletMetaCloudPB tablet_meta;
+ Versionstamp versionstamp;
+ ASSERT_EQ(reader.get_tablet_meta(tablet_id, &tablet_meta, &versionstamp),
TxnErrorCode::TXN_OK);
+ EXPECT_EQ(tablet_meta.table_id(), 1001);
+
+ // Decoupling the deleted intermediate clone clears both lineage fields
and its mapping.
+ instance_b.clear_source_instance_id();
+ instance_b.clear_source_snapshot_id();
+ resource_mgr.refresh_instance(instance_ids_[1], instance_b);
+
+ std::string source_instance_id;
+ Versionstamp source_snapshot_version;
+ EXPECT_FALSE(resource_mgr.get_source_snapshot_info(instance_ids_[1],
&source_instance_id,
+
&source_snapshot_version));
+}
+
TEST_F(CloneChainReaderTest, GetTabletSchema) {
std::string instance_id = instance_ids_[2]; // C
Versionstamp snapshot_version = snapshot_versions_[2];
diff --git a/cloud/test/resource_test.cpp b/cloud/test/resource_test.cpp
index a1208dbfbac..206f563a20a 100644
--- a/cloud/test/resource_test.cpp
+++ b/cloud/test/resource_test.cpp
@@ -446,11 +446,11 @@ TEST(ResourceTest,
RefreshDeletedInstanceClearsRuntimeCaches) {
resource_mgr.refresh_instance(instance_id, instance);
nodes.clear();
- EXPECT_EQ(resource_mgr.get_node(cloud_unique_id, &nodes), "cloud_unique_id
not found");
- EXPECT_TRUE(nodes.empty());
- EXPECT_FALSE(resource_mgr.is_version_read_enabled(instance_id));
- EXPECT_FALSE(resource_mgr.get_source_snapshot_info(instance_id,
&source_instance_id,
-
&source_snapshot_version));
+ EXPECT_EQ(resource_mgr.get_node(cloud_unique_id, &nodes), "");
+ ASSERT_TRUE(!nodes.empty());
+ EXPECT_TRUE(resource_mgr.is_version_read_enabled(instance_id));
+ EXPECT_TRUE(resource_mgr.get_source_snapshot_info(instance_id,
&source_instance_id,
+
&source_snapshot_version));
}
// test add/drop cluster
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
