[GitHub] [iceberg] rymurr commented on a change in pull request #2295: Spark: expire catalog cache to avoid hive connection overflow

[GitBox <gi...@apache.org>]

rymurr commented on a change in pull request #2295:
URL: https://github.com/apache/iceberg/pull/2295#discussion_r595051241



##########
File path: spark2/src/main/java/org/apache/iceberg/spark/source/CustomCatalogs.java
##########
@@ -40,6 +43,12 @@
 
 public final class CustomCatalogs {
   private static final Cache<Pair<SparkSession, String>, Catalog> CATALOG_CACHE = Caffeine.newBuilder()
+      .expireAfterAccess(10, TimeUnit.MINUTES)

Review comment:
       I agree, I think the timed expiration is potentially worse here. If we expire by time then we will be evicting a catalog that is definitely still being used by a `SparkSession`. Whereas before it was _maybe_ being used. 
   
   I think that #2325 may actually be the right solution. It moves to a cross catalog pool and handles reclaiming client connections regardless of the underlying catalog. Then any catalogs in this cache will have 0 connections attached to them and can stay in the cache forever. @lcspinter do you agree?




----------------------------------------------------------------
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@iceberg.apache.org
For additional commands, e-mail: issues-help@iceberg.apache.org