Abhishek Rawat has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/24283 )

Change subject: IMPALA-14962: Query Profile Parser and Section Retrieval 
Interface
......................................................................


Patch Set 12:

(1 comment)

http://gerrit.cloudera.org:8080/#/c/24283/11/be/src/service/query-profile-parsing-tools.h
File be/src/service/query-profile-parsing-tools.h:

http://gerrit.cloudera.org:8080/#/c/24283/11/be/src/service/query-profile-parsing-tools.h@94
PS11, Line 94:   rapidjson::Value GetNode(std::string_view node_id, 
std::string_view fragment_id,
I think we should call it GetPlanNode? The LLM might get confused between 
physical nodes and plan nodes. We also have GetPerNodeProfiles() which is the 
physical node level view and should probably remain as is since we use a lot of 
other per node keys in the profile to refer to node level metadata.

We should probably have a way for the LLM to request all of these:
* Average Plan Node section for a given Plan Node id
   ```
    HASH_JOIN_NODE (id=45) [132 instances]:(Total: 32s323ms, non-child: 
115.150ms, % non-child: 0.30%)
    ......
    ```

* All Plan Node sections for a given Plan Node id. It's important to include 
node (hostname) and fragment instance context along with Plan Node
   ```
      Instance 5d4385475b566934:1778e01d00000126 
(host=impala-executor-000-8.impala-executor.impala-1767912832-6hh7.svc.cluster.local:27010):
      ....
      ....
      HDFS_SCAN_NODE (id=45):
          Table Name: tpcds_3000_string_parquet_managed.store_sales
      ....

     Instance 5d4385475b566934:1778e01d000001cb 
(host=impala-executor-000-1.impala-executor.impala-1767912832-6hh7.svc.cluster.local:27010)
      .....
      .....
     HDFS_SCAN_NODE (id=45):
          Table Name: tpcds_3000_string_parquet_managed.store_sales

* Specific Plan Node section for a given Plan Node id. It's important to have 
node (hostname) and fragment instance context when evaluating specific Plan Node
   ```
    Instance 5d4385475b566934:1778e01d00000123 
(host=impala-executor-000-8.impala-executor.impala-1767912832-6hh7.svc.cluster.local:27010)
    ....
    HDFS_SCAN_NODE (id=45):
          Table Name: tpcds_3000_string_parquet_managed.store_sales
   ```



--
To view, visit http://gerrit.cloudera.org:8080/24283
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I4d9c9e735aa089bb243b07af421553002a465a88
Gerrit-Change-Number: 24283
Gerrit-PatchSet: 12
Gerrit-Owner: Gokul Kolady <[email protected]>
Gerrit-Reviewer: Abhishek Rawat <[email protected]>
Gerrit-Reviewer: Gokul Kolady <[email protected]>
Gerrit-Reviewer: Impala Public Jenkins <[email protected]>
Gerrit-Reviewer: Jason Fehr <[email protected]>
Gerrit-Reviewer: Yida Wu <[email protected]>
Gerrit-Comment-Date: Mon, 18 May 2026 23:18:34 +0000
Gerrit-HasComments: Yes

Reply via email to