wchevreuil commented on code in PR #8218:
URL: https://github.com/apache/hbase/pull/8218#discussion_r3225386150
##########
hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/CacheAwareLoadBalancer.java:
##########
@@ -61,6 +62,30 @@ public class CacheAwareLoadBalancer extends
StochasticLoadBalancer {
"hbase.master.balancer.stochastic.throttling.cacheRatio";
public static final float CACHE_RATIO_THRESHOLD_DEFAULT = 0.8f;
+ /**
+ * Below this cache ratio on the current host, a move may be considered for
the free-space
+ * heuristic.
+ */
+ public static final String LOW_CACHE_RATIO_FOR_RELOCATION_KEY =
+ "hbase.master.balancer.cacheaware.lowCacheRatioThreshold";
+ public static final float LOW_CACHE_RATIO_FOR_RELOCATION_DEFAULT = 0.35f;
+
+ /**
+ * Optimistic region cache ratio assumed for cost purposes when a better
host has free cache space
+ * (actual warmup is not modeled).
+ */
+ public static final String POTENTIAL_CACHE_RATIO_AFTER_MOVE_KEY =
+ "hbase.master.balancer.cacheaware.potentialCacheRatioAfterMove";
+ public static final float POTENTIAL_CACHE_RATIO_AFTER_MOVE_DEFAULT = 0.95f;
Review Comment:
For potentialCacheRatioAfterMove, I haven't formally measured. For root dirs
on cloud storage, I've seen performance degradation starts to be noticeable
when cache ratio falls below 90%.
For minFreeCacheSpaceFactor, I thought originally of only allow moving if
cache free space can accommodate the whole region, but decided to make this
configurable for flexibility.
--
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]
