[FLINK-34199] Add documentation.

docs/content.zh/docs/ops/traces.md

@@ -97,7 +97,7 @@ Flink reports a single span trace for the whole checkpoint and job initializatio
   </thead>
   <tbody>
     <tr>
-      <th rowspan="16">org.apache.flink.</br>runtime.checkpoint.</br>CheckpointStatsTracker</th>
+      <th rowspan="18">org.apache.flink.</br>runtime.checkpoint.</br>CheckpointStatsTracker</th>
       <th rowspan="6"><strong>Checkpoint</strong></th>
       <td>startTs</td>
       <td>Timestamp when the checkpoint has started.</td>
@@ -123,7 +123,7 @@ Flink reports a single span trace for the whole checkpoint and job initializatio
       <td>What was the state of this checkpoint: FAILED or COMPLETED.</td>
     </tr>
     <tr>
-      <th rowspan="10"><strong>JobInitialization</strong></th>
+      <th rowspan="12"><strong>JobInitialization</strong></th>
       <td>startTs</td>
       <td>Timestamp when the job initialization has started.</td>
     </tr>
@@ -157,7 +157,11 @@ Flink reports a single span trace for the whole checkpoint and job initializatio
     </tr>
     <tr>
       <td>(Max/Sum)DownloadStateDurationMs<br><br>(optional - currently only supported by RocksDB Incremental)</td>
-      <td>The aggregated (max and sum) across all subtasks duration of downloading state files from the DFS.</td>
+      <td>The aggregated (max and sum) duration across all subtasks of downloading state files from the DFS.</td>
+    </tr>
+    <tr>
+      <td>(Max/Sum)RestoreStateDurationMs<br><br>(optional - currently only supported by RocksDB Incremental)</td>
+      <td>The aggregated (max and sum) duration across all subtasks of restoring the state backend from fully localized state, i.e. after all remote state was downloaded.</td>
     </tr>
     <tr>
       <td>(Max/Sum)RestoredStateSizeBytes.[location]</td>
@@ -167,6 +171,10 @@ Flink reports a single span trace for the whole checkpoint and job initializatio
         REMOTE,
         UNKNOWN.</td>
     </tr>
+    <tr>
+      <td>(Max/Sum)RestoreAsyncCompactionDurationMs<br><br>(optional - currently only supported by RocksDB Incremental)</td>
+      <td>The aggregated (max and sum) duration across all subtasks for async compaction after incremental restore.</td>
+    </tr>
   </tbody>
 </table>
 

docs/content/docs/ops/traces.md

@@ -97,7 +97,7 @@ Flink reports a single span trace for the whole checkpoint and job initializatio
   </thead>
   <tbody>
     <tr>
-      <th rowspan="16">org.apache.flink.</br>runtime.checkpoint.</br>CheckpointStatsTracker</th>
+      <th rowspan="18">org.apache.flink.</br>runtime.checkpoint.</br>CheckpointStatsTracker</th>
       <th rowspan="6"><strong>Checkpoint</strong></th>
       <td>startTs</td>
       <td>Timestamp when the checkpoint has started.</td>
@@ -123,7 +123,7 @@ Flink reports a single span trace for the whole checkpoint and job initializatio
       <td>What was the state of this checkpoint: FAILED or COMPLETED.</td>
     </tr>
     <tr>
-      <th rowspan="10"><strong>JobInitialization</strong></th>
+      <th rowspan="12"><strong>JobInitialization</strong></th>
       <td>startTs</td>
       <td>Timestamp when the job initialization has started.</td>
     </tr>
@@ -157,7 +157,11 @@ Flink reports a single span trace for the whole checkpoint and job initializatio
     </tr>
     <tr>
       <td>(Max/Sum)DownloadStateDurationMs<br><br>(optional - currently only supported by RocksDB Incremental)</td>
-      <td>The aggregated (max and sum) across all subtasks duration of downloading state files from the DFS.</td>
+      <td>The aggregated (max and sum) duration across all subtasks of downloading state files from the DFS.</td>
+    </tr>
+    <tr>
+      <td>(Max/Sum)RestoreStateDurationMs<br><br>(optional - currently only supported by RocksDB Incremental)</td>
+      <td>The aggregated (max and sum) duration across all subtasks of restoring the state backend from fully localized state, i.e. after all remote state was downloaded.</td>
     </tr>
     <tr>
       <td>(Max/Sum)RestoredStateSizeBytes.[location]</td>
@@ -167,6 +171,10 @@ Flink reports a single span trace for the whole checkpoint and job initializatio
         REMOTE,
         UNKNOWN.</td>
     </tr>
+    <tr>
+      <td>(Max/Sum)RestoreAsyncCompactionDurationMs<br><br>(optional - currently only supported by RocksDB Incremental)</td>
+      <td>The aggregated (max and sum) duration across all subtasks for async compaction after incremental restore.</td>
+    </tr>
   </tbody>
 </table>
 

docs/layouts/shortcodes/generated/rocksdb_configurable_configuration.html

@@ -74,6 +74,12 @@
             <td>Integer</td>
             <td>The maximum number of open files (per stateful operator) that can be used by the DB, '-1' means no limit. The default value is '-1'.</td>
         </tr>
+        <tr>
+            <td><h5>state.backend.rocksdb.incremental-restore-async-compact-after-rescale</h5></td>
+            <td style="word-wrap: break-word;">false</td>
+            <td>Boolean</td>
+            <td>If true, an async compaction of RocksDB is started after every restore after which we detect keys (including tombstones) in the database that are outside the key-groups range of the backend.</td>
+        </tr>
         <tr>
             <td><h5>state.backend.rocksdb.log.dir</h5></td>
             <td style="word-wrap: break-word;">(none)</td>
@@ -116,6 +122,12 @@
             <td>Boolean</td>
             <td>If true, every newly created SST file will contain a Bloom filter. It is disabled by default.</td>
         </tr>
+        <tr>
+            <td><h5>state.backend.rocksdb.use-ingest-db-restore-mode</h5></td>
+            <td style="word-wrap: break-word;">false</td>
+            <td>Boolean</td>
+            <td>A recovery mode that directly clips and ingests multiple DBs during state recovery if the keys in the SST files does not exceed the declared key-group range.</td>
+        </tr>
         <tr>
             <td><h5>state.backend.rocksdb.write-batch-size</h5></td>
             <td style="word-wrap: break-word;">2 mb</td>