kadirozde commented on code in PR #1896:
URL: https://github.com/apache/phoenix/pull/1896#discussion_r1616049951
##########
phoenix-core-server/src/main/java/org/apache/phoenix/coprocessor/UngroupedAggregateRegionObserver.java:
##########
@@ -581,92 +601,135 @@ private boolean areMutationsInSameTable(Table
targetHTable, Region region) {
region.getTableDescriptor().getTableName().getName()) == 0);
}
+ @Override
+ public InternalScanner
preFlush(ObserverContext<RegionCoprocessorEnvironment> c, Store store,
+ InternalScanner scanner, FlushLifeCycleTracker tracker) throws
IOException {
+ if (!isPhoenixTableTTLEnabled(c.getEnvironment().getConfiguration())) {
+ return scanner;
+ } else {
+ return User.runAsLoginUser(new
PrivilegedExceptionAction<InternalScanner>() {
+ @Override public InternalScanner run() throws Exception {
+ String tableName =
c.getEnvironment().getRegion().getRegionInfo().getTable()
+ .getNameAsString();
+ long maxLookbackInMillis =
+
UngroupedAggregateRegionObserver.getMaxLookbackInMillis(
Review Comment:
Yes, you are right that this solution here is eventual consistent. The
precompact is called for minor and major compaction. Yes, major compaction is
done periodically typically once a week. However, the compaction policy may
decide to skip it. Minor compaction is supposed to run whenever the number of
HFiles exceeds the configured threshold. If the eventual consistency is fine
for data table regions, then eventual consistency between data table region and
index from the max lookback perspective should be also fine. However, I think
we need a better solution than this such that we need to populate and
invalidate this cache whenever the max lookback config changes. The solution
here is a stopgap. Hopefully, we will implement a better solution before 5.3.
Please note we can continue to configure "phoenix.max.lookback.age.seconds" at
the table level to using the Phoenix alter table command to get much better
eventual consistency meanwhile.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
