-
Type:
Task
-
Resolution: Won't Fix
-
Priority:
Minor - P4
-
None
-
Affects Version/s: None
-
Component/s: None
-
None
WT-4732 Introduced a regression into the million collection test, the purpose of this ticket is to determine why it caused a regression and if it can be resolved or if that regression can be flagged as works as intended.
The million collection test expected a read throughput of 7500 op/s however that fell to approximately 6300 failing the test.
[2019/07/02 10:50:41.477] java -jar ../POCDriver/bin/POCDriver.jar -i 0 -u 100 -k 0 -q 10000 -z 549 -d 300 -y 72745 --collectionKeyMax 549 -o ../results/out.csv -t 64 -b 1 [2019/07/02 10:55:43.453] Run completed avg latency with 72745 collections is 30.3910063679ms/op and throughput of 6319ops/sec [2019/07/02 10:55:43.453] The run completed with statistics below: [2019/07/02 10:55:43.453] - avg response time (latency): 30.3910063679 (ms/op), avg throughput: 6319 (ops/sec) [2019/07/02 10:55:43.453] - total run time: 9055.63679218, configured total_runtime: 86400 [2019/07/02 10:55:43.453] - num of collections: 72745, configured num_collections: 75000 [2019/07/02 10:55:43.453] - threads: 64, configured num_threads: 64 [2019/07/02 10:55:43.453] Average throughput is less than 7500.0 - test failed!
Additionally a separate ticket WT-4494 introduced a regression around the same time so its unclear exactly how much can be attributed to WT-4732. This is being tracked in WT-5056. The test was modified to account for this regression and pass anyway, ideally we'd like to undo that modification as such if the regression can be fixed, it should include removing that modification.
See this jira for full details: WT-4905.
- related to
-
WT-4732 Remove the bias from the eviction policy on the dhandle order
-
- Closed
-