[
https://issues.apache.org/jira/browse/IGNITE-25369?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Konstantin Orlov updated IGNITE-25369:
--------------------------------------
Description:
Apache Ignite is distributed system. It has an ability to have several copies
of the data named replicas. Therefore, to execute a query, original plan is
split on number of fragments, and every fragment is mapped on subset of nodes.
It would be beneficial to provide an insight on which nodes will be used to
execute which part of the query.
For this, let's introduce to command {{EXPLAIN MAPPING FOR <queryOrDml>}}.
Below is suggested output of the command:
{code}
EXPLAIN MAPPING FOR SELECT
U.UserName, P.ProductName, R.ReviewText, R.Rating
FROM Users U, Reviews R, Products P
WHERE U.UserID = R.UserID
AND R.ProductID = P.ProductID
AND P.ProductName = 'Product_' || ?::varchar;
Fragment#0 root
executionNodes: [ijot_n_3344]
plan:
Project
fields: [USERNAME, PRODUCTNAME, REVIEWTEXT, RATING]
exprs: [USERNAME, PRODUCTNAME, REVIEWTEXT, RATING]
est: (rows=1)
HashJoin
condition: =(USERID0, USERID)
joinType: inner
est: (rows=1)
HashJoin
condition: =(PRODUCTID, PRODUCTID0)
joinType: inner
est: (rows=1)
Receiver
rowType: RecordType(INTEGER PRODUCTID, INTEGER USERID, VARCHAR(65536)
REVIEWTEXT, INTEGER RATING)
exchangeId: 1
sourceFragmentId: 1
est: (rows=1)
Receiver
rowType: RecordType(INTEGER PRODUCTID, VARCHAR(100) PRODUCTNAME)
exchangeId: 2
sourceFragmentId: 2
est: (rows=1)
Receiver
rowType: RecordType(INTEGER USERID, VARCHAR(100) USERNAME)
exchangeId: 3
sourceFragmentId: 3
est: (rows=1)
Fragment#1
targetFragment: 0
executionNodes: [ijot_n_3344, ijot_n_3345, ijot_n_3346]
plan:
Sender
exchangeId: 1
targetFragmentId: 0
distribution: single
est: (rows=50000)
TableScan
table: PUBLIC.REVIEWS
fields: [PRODUCTID, USERID, REVIEWTEXT, RATING]
sourceId: 6
est: (rows=50000)
Fragment#2
targetFragment: 0
executionNodes: [ijot_n_3344, ijot_n_3345, ijot_n_3346]
plan:
Sender
exchangeId: 2
targetFragmentId: 0
distribution: single
est: (rows=1665)
TableScan
table: PUBLIC.PRODUCTS
filters: =(PRODUCTNAME, ||(_UTF-8'Product_', CAST(?0):VARCHAR CHARACTER
SET "UTF-8"))
fields: [PRODUCTID, PRODUCTNAME]
sourceId: 5
est: (rows=1665)
Fragment#3
targetFragment: 0
executionNodes: [ijot_n_3344, ijot_n_3345, ijot_n_3346]
plan:
Sender
exchangeId: 3
targetFragmentId: 0
distribution: single
est: (rows=10000)
TableScan
table: PUBLIC.USERS
fields: [USERID, USERNAME]
sourceId: 4
est: (rows=10000)
{code}
h3. Implementation Notes
As a first step, it worth to reuse
{{org.apache.ignite.internal.sql.engine.exec.mapping.FragmentPrinter}} as is,
and adjust output in follow up ticket.
was:
Apache Ignite is distributed system. It has an ability to have several copies
of the data named replicas. Therefore, to execute a query, original plan is
split on number of fragments, and every fragment is mapped on subset of nodes.
It would be beneficial to provide an insight on which nodes will be used to
execute which part of the query.
For this, let's introduce to command {{EXPLAIN MAPPING FOR <queryOrDml>}}.
Below is suggested output of the command:
{code}
EXPLAIN MAPPING FOR SELECT
U.UserName, P.ProductName, R.ReviewText, R.Rating
FROM Users U, Reviews R, Products P
WHERE U.UserID = R.UserID
AND R.ProductID = P.ProductID
AND P.ProductName = 'Product_' || ?::varchar;
Fragment#0 root
executionNodes: [ijot_n_3344]
plan:
Project
fields: [USERNAME, PRODUCTNAME, REVIEWTEXT, RATING]
exprs: [USERNAME, PRODUCTNAME, REVIEWTEXT, RATING]
est: (rows=1)
HashJoin
condition: =(USERID0, USERID)
joinType: inner
est: (rows=1)
HashJoin
condition: =(PRODUCTID, PRODUCTID0)
joinType: inner
est: (rows=1)
Receiver
rowType: RecordType(INTEGER PRODUCTID, INTEGER USERID, VARCHAR(65536)
REVIEWTEXT, INTEGER RATING)
exchangeId: 1
sourceFragmentId: 1
est: (rows=1)
Receiver
rowType: RecordType(INTEGER PRODUCTID, VARCHAR(100) PRODUCTNAME)
exchangeId: 2
sourceFragmentId: 2
est: (rows=1)
Receiver
rowType: RecordType(INTEGER USERID, VARCHAR(100) USERNAME)
exchangeId: 3
sourceFragmentId: 3
est: (rows=1)
Fragment#1
targetFragment: 0
executionNodes: [ijot_n_3344, ijot_n_3345, ijot_n_3346]
plan:
Sender
exchangeId: 1
targetFragmentId: 0
distribution: single
est: (rows=50000)
TableScan
table: PUBLIC.REVIEWS
fields: [PRODUCTID, USERID, REVIEWTEXT, RATING]
sourceId: 6
est: (rows=50000)
Fragment#2
targetFragment: 0
executionNodes: [ijot_n_3344, ijot_n_3345, ijot_n_3346]
plan:
Sender
exchangeId: 2
targetFragmentId: 0
distribution: single
est: (rows=1665)
TableScan
table: PUBLIC.PRODUCTS
filters: =(PRODUCTNAME, ||(_UTF-8'Product_', CAST(?0):VARCHAR CHARACTER
SET "UTF-8"))
fields: [PRODUCTID, PRODUCTNAME]
sourceId: 5
est: (rows=1665)
Fragment#3
targetFragment: 0
executionNodes: [ijot_n_3344, ijot_n_3345, ijot_n_3346]
plan:
Sender
exchangeId: 3
targetFragmentId: 0
distribution: single
est: (rows=10000)
TableScan
table: PUBLIC.USERS
fields: [USERID, USERNAME]
sourceId: 4
est: (rows=10000)
{code}
> Sql. Introduce EXPLAIN MAPPING FOR command
> ------------------------------------------
>
> Key: IGNITE-25369
> URL: https://issues.apache.org/jira/browse/IGNITE-25369
> Project: Ignite
> Issue Type: Improvement
> Components: sql ai3
> Reporter: Konstantin Orlov
> Priority: Major
> Labels: ignite-3
>
> Apache Ignite is distributed system. It has an ability to have several copies
> of the data named replicas. Therefore, to execute a query, original plan is
> split on number of fragments, and every fragment is mapped on subset of
> nodes. It would be beneficial to provide an insight on which nodes will be
> used to execute which part of the query.
> For this, let's introduce to command {{EXPLAIN MAPPING FOR <queryOrDml>}}.
> Below is suggested output of the command:
> {code}
> EXPLAIN MAPPING FOR SELECT
> U.UserName, P.ProductName, R.ReviewText, R.Rating
> FROM Users U, Reviews R, Products P
> WHERE U.UserID = R.UserID
> AND R.ProductID = P.ProductID
> AND P.ProductName = 'Product_' || ?::varchar;
> Fragment#0 root
> executionNodes: [ijot_n_3344]
> plan:
> Project
> fields: [USERNAME, PRODUCTNAME, REVIEWTEXT, RATING]
> exprs: [USERNAME, PRODUCTNAME, REVIEWTEXT, RATING]
> est: (rows=1)
> HashJoin
> condition: =(USERID0, USERID)
> joinType: inner
> est: (rows=1)
> HashJoin
> condition: =(PRODUCTID, PRODUCTID0)
> joinType: inner
> est: (rows=1)
> Receiver
> rowType: RecordType(INTEGER PRODUCTID, INTEGER USERID,
> VARCHAR(65536) REVIEWTEXT, INTEGER RATING)
> exchangeId: 1
> sourceFragmentId: 1
> est: (rows=1)
> Receiver
> rowType: RecordType(INTEGER PRODUCTID, VARCHAR(100) PRODUCTNAME)
> exchangeId: 2
> sourceFragmentId: 2
> est: (rows=1)
> Receiver
> rowType: RecordType(INTEGER USERID, VARCHAR(100) USERNAME)
> exchangeId: 3
> sourceFragmentId: 3
> est: (rows=1)
> Fragment#1
> targetFragment: 0
> executionNodes: [ijot_n_3344, ijot_n_3345, ijot_n_3346]
> plan:
> Sender
> exchangeId: 1
> targetFragmentId: 0
> distribution: single
> est: (rows=50000)
> TableScan
> table: PUBLIC.REVIEWS
> fields: [PRODUCTID, USERID, REVIEWTEXT, RATING]
> sourceId: 6
> est: (rows=50000)
> Fragment#2
> targetFragment: 0
> executionNodes: [ijot_n_3344, ijot_n_3345, ijot_n_3346]
> plan:
> Sender
> exchangeId: 2
> targetFragmentId: 0
> distribution: single
> est: (rows=1665)
> TableScan
> table: PUBLIC.PRODUCTS
> filters: =(PRODUCTNAME, ||(_UTF-8'Product_', CAST(?0):VARCHAR CHARACTER
> SET "UTF-8"))
> fields: [PRODUCTID, PRODUCTNAME]
> sourceId: 5
> est: (rows=1665)
> Fragment#3
> targetFragment: 0
> executionNodes: [ijot_n_3344, ijot_n_3345, ijot_n_3346]
> plan:
> Sender
> exchangeId: 3
> targetFragmentId: 0
> distribution: single
> est: (rows=10000)
> TableScan
> table: PUBLIC.USERS
> fields: [USERID, USERNAME]
> sourceId: 4
> est: (rows=10000)
> {code}
> h3. Implementation Notes
> As a first step, it worth to reuse
> {{org.apache.ignite.internal.sql.engine.exec.mapping.FragmentPrinter}} as is,
> and adjust output in follow up ticket.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
