Conversation
Jackie-Jiang
left a comment
There was a problem hiding this comment.
There is one corner case not handled. When a record is deleted, we might want to track its timestamp to guarantee consistency for late event. This could cause much bigger map size thus revoke the benefit of deletion. We can further discuss the tradeoffs
...al/src/main/java/org/apache/pinot/segment/local/indexsegment/mutable/MutableSegmentImpl.java
Outdated
Show resolved
Hide resolved
.../java/org/apache/pinot/segment/local/upsert/ConcurrentMapPartitionUpsertMetadataManager.java
Show resolved
Hide resolved
.../java/org/apache/pinot/segment/local/upsert/ConcurrentMapPartitionUpsertMetadataManager.java
Outdated
Show resolved
Hide resolved
.../java/org/apache/pinot/segment/local/upsert/ConcurrentMapPartitionUpsertMetadataManager.java
Show resolved
Hide resolved
|
can we see all the configs related to upsert? lets try to keep the config keys intuitive and follow some pattern |
I think we can keep that in the state for now. The current ask is to just support deleting the record afaik (i.e. to mark it as invalid). Reducing the state size can be thought of as an optimization. |
Definitely. This is just a Draft PR. I thought about using the existing |
Codecov Report
@@ Coverage Diff @@
## master #10035 +/- ##
============================================
- Coverage 70.42% 70.34% -0.08%
- Complexity 5691 5697 +6
============================================
Files 1994 1994
Lines 107508 107579 +71
Branches 16340 16353 +13
============================================
- Hits 75709 75679 -30
- Misses 26502 26599 +97
- Partials 5297 5301 +4
Flags with carried forward coverage won't be shown. Click here to find out more.
📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
Use two new configs -