[ 
https://issues.apache.org/jira/browse/HDDS-15059?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18134002#comment-18134002
 ] 

Aleksei Ieshin edited comment on HDDS-15059 at 10/29/26, 6:04 AM:
------------------------------------------------------------------

The read path has the same problem this change handles for the opt-in OM-side 
write sort. {{KeyManagerImpl.sortDatanodes(String, List<OmKeyInfo>)}} still 
passes the RPC-deserialized {{pipeline.getNodes()}} straight to 
{{{}sortByDistanceCost{}}}, without mapping them through the cluster map like 
{{sortByClusterMapDistance}} does.

So {{getDistanceCost}} returns {{Integer.MAX_VALUE }}for every node except the 
client, and topology-aware reads are effectively "local host first, then 
random". We reproduced it with a unit test (same-rack nodes first in 76/1000 
sorts against 1000/1000 with linked nodes). Filed as HDDS-17555.


was (Author: JIRAUSER311863):
The read path has the same problem this change handles for the opt-in OM-side 
write sort. {{KeyManagerImpl.sortDatanodes(String, List<OmKeyInfo>)}} still 
passes the RPC-deserialized {{pipeline.getNodes()}} straight to 
{{{}sortByDistanceCost{}}}, without mapping them through the cluster map like 
{{sortByClusterMapDistance}} does.

So {{getDistanceCost}} returns {{Integer.MAX_VALUE}} 
for every node except the client, and topology-aware reads are effectively 
"local host first, then random". We reproduced it with a unit test (same-rack 
nodes first in 76/1000 sorts against 1000/1000 with linked nodes). Filed as 
HDDS-17555.

> Shift streaming write sortDatanodes logic to OM
> -----------------------------------------------
>
>                 Key: HDDS-15059
>                 URL: https://issues.apache.org/jira/browse/HDDS-15059
>             Project: Apache Ozone
>          Issue Type: Improvement
>            Reporter: Ivan Andika
>            Assignee: Chi-Hsuan Huang
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: 2.3.0
>
>
> Since HDDS-9343 allows OM to fetch clusterMap from SCM and store in OM 
> locally. We can do the sorting on the OM instead of SCM to reduce SCM RPC 
> processing time.



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to