Repository navigation
[core][raylet] Fix spill_manager_objects_bytes reports restored object count instead of restored bytes - #65013
Conversation
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.
Reviewed by Cursor Bugbot for commit 540d2d3. Configure here.
|
Hi, @edoakes @Sparks0219 Could you help take a look if you had the free time ? Thank you very much! |
There was a problem hiding this comment.
Code Review
This pull request corrects a metric recording bug in LocalObjectManager where restored_objects_total_ was incorrectly used instead of restored_bytes_total_ for recording restored bytes. It also adds a unit test to verify this behavior. The review feedback points out an accidental indentation change in the modified line that should be restored to maintain code consistency.
c51e2d3 to
24c4218
Compare
|
Hi @RocMarshal, please fix the lint. |
44f5eba to
093abbc
Compare
|
Thanks @rueian for the reminder. |
058624a to
4f025d7
Compare
…t count instead of restored bytes Signed-off-by: Yuepeng Pan <panyuepeng@apache.org>
|
Thanks @rueian ☕ |
…t count instead of restored bytes (ray-project#65013) ## Description - What problem does this solve? The `spill_manager_objects_bytes` metric appears to report the wrong value for `State="Restored"`. In `LocalObjectManager::RecordMetrics()`, the restored sample for `spill_manager_objects_bytes` is recorded with `restored_objects_total_`: ```cpp spill_manager_metrics_.spill_manager_objects_bytes_gauge.Record( restored_objects_total_, {{"State", "Restored"}}); ``` However, this metric is documented as a byte-size metric: name = "spill_manager_objects_bytes" description = "Byte size of local objects broken per state ..." - How does this PR solve it? correct the new value of the target metrics . - Any relevant context for reviewers such as: ray-project#65012 - Why is the problem important to solve? IIUC, It's a but for metrics on the corresponding value. - Why was this approach chosen over others? direct & simple. ## Related issues Fixes ray-project#65012 ## Additional information N.A Signed-off-by: Yuepeng Pan <panyuepeng@apache.org>
…t count instead of restored bytes (ray-project#65013) ## Description - What problem does this solve? The `spill_manager_objects_bytes` metric appears to report the wrong value for `State="Restored"`. In `LocalObjectManager::RecordMetrics()`, the restored sample for `spill_manager_objects_bytes` is recorded with `restored_objects_total_`: ```cpp spill_manager_metrics_.spill_manager_objects_bytes_gauge.Record( restored_objects_total_, {{"State", "Restored"}}); ``` However, this metric is documented as a byte-size metric: name = "spill_manager_objects_bytes" description = "Byte size of local objects broken per state ..." - How does this PR solve it? correct the new value of the target metrics . - Any relevant context for reviewers such as: ray-project#65012 - Why is the problem important to solve? IIUC, It's a but for metrics on the corresponding value. - Why was this approach chosen over others? direct & simple. ## Related issues Fixes ray-project#65012 ## Additional information N.A Signed-off-by: Yuepeng Pan <panyuepeng@apache.org> Signed-off-by: 400Ping <jiekaichang@apache.org>
…t count instead of restored bytes (ray-project#65013) ## Description - What problem does this solve? The `spill_manager_objects_bytes` metric appears to report the wrong value for `State="Restored"`. In `LocalObjectManager::RecordMetrics()`, the restored sample for `spill_manager_objects_bytes` is recorded with `restored_objects_total_`: ```cpp spill_manager_metrics_.spill_manager_objects_bytes_gauge.Record( restored_objects_total_, {{"State", "Restored"}}); ``` However, this metric is documented as a byte-size metric: name = "spill_manager_objects_bytes" description = "Byte size of local objects broken per state ..." - How does this PR solve it? correct the new value of the target metrics . - Any relevant context for reviewers such as: ray-project#65012 - Why is the problem important to solve? IIUC, It's a but for metrics on the corresponding value. - Why was this approach chosen over others? direct & simple. ## Related issues Fixes ray-project#65012 ## Additional information N.A Signed-off-by: Yuepeng Pan <panyuepeng@apache.org>

Description
What problem does this solve?
The
spill_manager_objects_bytesmetric appears to report the wrong value forState="Restored".In
LocalObjectManager::RecordMetrics(), the restored sample forspill_manager_objects_bytesis recorded withrestored_objects_total_:spill_manager_metrics_.spill_manager_objects_bytes_gauge.Record( restored_objects_total_, {{"State", "Restored"}});However, this metric is documented as a byte-size metric:
name = "spill_manager_objects_bytes"
description = "Byte size of local objects broken per state ..."
How does this PR solve it?
correct the new value of the target metrics .
Any relevant context for reviewers such as:
[core][raylet] spill_manager_objects_bytes reports restored object count instead of restored bytes #65012
Why is the problem important to solve?
IIUC, It's a but for metrics on the corresponding value.
Why was this approach chosen over others?
direct & simple.
Related issues
Fixes #65012
Additional information
N.A