Skip to content

Conversation

@maobaolong
Copy link
Member

@maobaolong maobaolong commented Nov 26, 2024

What changes were proposed in this pull request?

  • Add include key filter before report extraProperties
  • Related docs

Why are the changes needed?

  • Add include key filter
  • Add documents.

Does this PR introduce any user-facing change?

No.

How was this patch tested?

Tested on our test cluster.

  • Client conf
--conf spark.shuffle.manager=org.apache.spark.shuffle.DelegationRssShuffleManager --conf spark.rss.access.id="access.id"
--conf spark.rss.client.reportIncludeProperties="spark.test.key,test2.key,app.id,yarn.queue" --conf spark.rss.client.reportExcludeProperties="app.id"
  • Coordinator_rpc_audit.log
[2024-11-26 21:14:45.632] cmd=accessCluster	statusCode=ACCESS_DENIED	from=/XXX:34733	executionTimeUs=276(<1s)	appId=N/A	args{accessInfo=AccessInfo{accessId='rss-test-mbl-003-bannedid', user= tdwadmin, tags=[ss_v5], extraProperties={access_info_required_shuffle_nodes_num=-1, yarn.queue=g_teg_tdw_operlog-offline, spark.test.key=123_456}}}

@maobaolong
Copy link
Member Author

@jerqi Thanks for the review of #2254, I added the related document and add white list config to improve this feature.

@github-actions
Copy link

github-actions bot commented Nov 26, 2024

Test Results

 2 966 files  ± 0   2 966 suites  ±0   6h 38m 44s ⏱️ + 11m 24s
 1 100 tests + 4   1 098 ✅ + 4   2 💤 ±0  0 ❌ ±0 
13 774 runs  +39  13 744 ✅ +39  30 💤 ±0  0 ❌ ±0 

Results for commit 1de03b7. ± Comparison against base commit db611f4.

♻️ This comment has been updated with latest results.

List<String> includeProperties =
rssConf.get(RssClientConf.RSS_CLIENT_REPORT_INCLUDE_PROPERTIES);
rssConf.getAll().stream()
.filter(entry -> includeProperties.isEmpty() || includeProperties.contains(entry.getKey()))
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why do we judge the includeProperties.isEmpty()?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Empty means include all, that means include config is disabled, it reference what hdfs treat include/excludes datanode list

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Empty should mean including empty.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So how to express including all?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If we have include list, you can't express including all.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Well, if I set the include option withNoDefaults, do you think it can make sense if we do not set this include option stand for include all?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's ok.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done

@maobaolong
Copy link
Member Author

@jerqi Sorry for the late reply, it really busy these days, addressed the suggestion from you PTAL.

@jerqi jerqi changed the title [#2254][FOLLOWUP] improvement(spark-client):Add include key filter before report extraProperties [#2254][FOLLOWUP] improvement(spark):Add include key filter before report extraProperties Dec 3, 2024
| <client_type>.rss.client.rpc.netty.maxOrder | 3 | The value of maxOrder for PooledByteBufAllocator when using gRPC internal Netty on the client-side. This configuration will only take effect when rss.rpc.server.type is set to GRPC_NETTY. |
| <client_type>.rss.client.rpc.netty.smallCacheSize | 1024 | The value of smallCacheSize for PooledByteBufAllocator when using gRPC internal Netty on the client-side. This configuration will only take effect when rss.rpc.server.type is set to GRPC_NETTY. |
| <client_type>.rss.client.blockIdManagerClass | - | The block id manager class of server for this application, the implementation of this interface to manage the shuffle block ids |
| <client_type>.rss.client.reportExcludeProperties | - | the report exclude properties could be configured by this option |
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is the document right about reportExcludeProperties and reportIncludeProperties?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done

@maobaolong maobaolong requested a review from jerqi January 3, 2025 03:35
@maobaolong
Copy link
Member Author

@jerqi I updated the document of those new configure keys. PTAL

Copy link
Contributor

@jerqi jerqi left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM.

@maobaolong
Copy link
Member Author

@jerqi Thanks for your review! Merged.

@maobaolong maobaolong merged commit fd27a08 into apache:master Jan 4, 2025
43 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants