退订
原始邮件
发件人：guangyu05 <[email protected]>
发件时间：2026年10月8日 13:02
收件人：user <[email protected]>
抄送：user <[email protected]>
主题：回复：[SURVEY] Do you use Apache Solr with Flink?
退订
---- 回复的原邮件 ----
发件人
Amrit Sarkar<[email protected]>
发送日期
2026年10月06日 13:32
收件人
user<[email protected]>
主题
[SURVEY] Do you use Apache Solr with Flink?
Hi all,
I'm trying to find out how many people use Flink together with Apache Solr to
decide whether an official Solr connector is worth building. There's been an
open request since 2018 (https://issues.apache.org/jira/browse/FLINK-8486), but
no upstream connector exists. From what I can find, teams write their own SolrJ
code or use third-party connectors such as ChunJun's.
For context, I've worked on Flink for about 5 years. I have seen teams write to
Solr from Flink with their own sink.
If you move data between Flink and Solr, or have tried and given up, it would
help a lot if you could answer any of these:
1. How do you do it today? Custom SolrJ code in a sink or window function,
ChunJun, Flink to Kafka to Kafka Connect, or something else?
2. What do you need: sink, bounded source, lookup join, or a mix?
3. DataStream API or SQL/Table API?
4. Solr version, SolrCloud or standalone, and auth (basic, Kerberos, TLS)?
5. Roughly what scale: docs per second or per day, number of collections?
6. What has been painful? Upgrading Solr or SolrJ, retries and error handling,
deletes, commits, dependency conflicts?
7. Would you switch to an official connector if one existed, and would you be
willing to help test or maintain it?
Short or partial answers are fine. Reply on the list or to me directly if you'd
rather not share details publicly. I'll post a summary back to this thread in a
couple of weeks.
Thanks,
Amrit Sarkar
