Aleksandr Efimov has posted comments on this change. ( http://gerrit.cloudera.org:8080/24822 )
Change subject: IMPALA-14598: Store HBO cache in Redis/Valkey ...................................................................... Patch Set 2: (2 comments) http://gerrit.cloudera.org:8080/#/c/24822/2/fe/src/main/java/org/apache/impala/service/HistoricalStats.java File fe/src/main/java/org/apache/impala/service/HistoricalStats.java: http://gerrit.cloudera.org:8080/#/c/24822/2/fe/src/main/java/org/apache/impala/service/HistoricalStats.java@141 PS2, Line 141: (HistoricalStatsValue<TPlanNodeRun>) cacheBackend_.getIfPresent( > This GET-modify-SET is not atomic across coordinators, so concurrent writer Could we make this update atomic now? Coordinators can still overwrite each other's runs; batching won't prevent that. http://gerrit.cloudera.org:8080/#/c/24822/2/fe/src/main/java/org/apache/impala/service/RedisCacheBackend.java File fe/src/main/java/org/apache/impala/service/RedisCacheBackend.java: http://gerrit.cloudera.org:8080/#/c/24822/2/fe/src/main/java/org/apache/impala/service/RedisCacheBackend.java@81 PS2, Line 81: poolConfig.setMaxTotal(maxConnections); > Redis GET runs in the planner path. A slow server can consume the timeout p Timeouts still add up per node and strategy. Could we bound total HBO lookup time per query and test multi-node planning against a slow server? -- To view, visit http://gerrit.cloudera.org:8080/24822 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: I43a171bcd436f57bcff14ceaaaa98c0f7dcec769 Gerrit-Change-Number: 24822 Gerrit-PatchSet: 2 Gerrit-Owner: Quanlong Huang <[email protected]> Gerrit-Reviewer: Aleksandr Efimov <[email protected]> Gerrit-Reviewer: Impala Public Jenkins <[email protected]> Gerrit-Reviewer: Quanlong Huang <[email protected]> Gerrit-Comment-Date: Tue, 15 Sep 2026 05:44:27 +0000 Gerrit-HasComments: Yes
