You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
[ML] Refactor AutodetectMemoryLimitIT to use memory limit constants and improve model size assertions (#135526)
Previously, the upper bound for model memory checks was set in absolute terms, which is not easy to understand and is brittle. I adjusted the assertions to ensure that the memory usage does not exceed 5% of the memory limit. Additionally, on Linux, we now report the process size (see #131981), which includes approximately 20 MB of native code overhead. I made handling this overhead more explicit.
More details:
Removed muted tests for testManyDistinctOverFields and testTooManyByAndOverFields.
Introduced constants for memory limits in AutodetectMemoryLimitIT.
Updated assertions to check effective model size against calculated limits.
Closes#132308Closes#132310Closes#132611
Copy file name to clipboardExpand all lines: x-pack/plugin/ml/qa/native-multi-node-tests/src/javaRestTest/java/org/elasticsearch/xpack/ml/integration/AutodetectMemoryLimitIT.java
+12-16Lines changed: 12 additions & 16 deletions
Original file line number
Diff line number
Diff line change
@@ -26,8 +26,6 @@
26
26
27
27
importstaticorg.hamcrest.Matchers.anyOf;
28
28
importstaticorg.hamcrest.Matchers.equalTo;
29
-
importstaticorg.hamcrest.Matchers.greaterThan;
30
-
importstaticorg.hamcrest.Matchers.lessThan;
31
29
32
30
/**
33
31
* A set of tests that ensure we comply to the model memory limit
0 commit comments