chore: right-size comparison backends for 1M-keyspace YCSB runs - #34
Merged
Conversation
Keep CPU uniform (3) and give each backend a comparable 6Gi memory footprint so the YCSB core workloads (1M keyspace) run without OOM: - ZooKeeper: set heapSize to 3072. bitnami defaults the JVM heap to 1024MB regardless of the container limit, which the 1M keyspace overflows (Java OOM -> CrashLoopBackOff). Container raised to 6Gi for non-heap headroom. - oxia (all sizes): raise the pebble read cache from the 512MB default to 3GB (readCacheSizeMB), so oxia caches within its footprint -- comparable to zk's heap / etcd's page cache. Container 6Gi. - etcd: container 6Gi (from 4Gi). Signed-off-by: Matteo Merli <mmerli@apache.org>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Right-size the comparison backends so they survive the YCSB core workloads (1M keyspace), keeping CPU uniform (3) and giving each backend a comparable 6Gi memory footprint.
heapSize: 3072. bitnami defaults the JVM heap to 1024MB regardless of the container limit, which the 1M keyspace overflows → Java OOM →CrashLoopBackOff. Container bumped to 6Gi for non-heap (metaspace, netty direct buffers) + OS headroom.oxia,oxia-3/6/12): raise the pebble read cache from the 512MB default to 3GB (readCacheSizeMB: 3072), so oxia actually caches within its footprint — comparable to zk's heap / etcd's page cache. Container 6Gi.Verification
zk, etcd, and oxia-3/6/12 all completed the YCSB throughput comparison on a real cluster (zk previously OOM'd on heap; the others ran clean at the new sizes).