reduce sampling interval for perf to 100Hz #161
Merged
Conversation
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
Contributor
|
Just curious, is the OOM killer used because there is no swap? And does 'perf report' memory usage directly correlate to perf.data size? I would have expected the memory usage to be similar, no matter the perf.data size (seems unlikely perf would need to keep all of this data in memory). |
Contributor
Author
|
The swap part is quite possible. OpenShift/Kube/Atomic Enterprise run swapless by design. |
Contributor
Author
|
I guess I need to play around a bit to answer your question. But honestly I think we should take advantage of the 10x reduction in sample size and associated overhead/cpu usage of the tracing. I think 100Hz is enough, and would get us what we need for both short and long runs. |
Contributor
|
Sure, I support the default of 100 |
atheurer
added a commit
that referenced
this pull request
Feb 10, 2016
reduce sampling interval for perf to 100Hz
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.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.
We found some cases where when a longer run (1hour) was run on a system with 32GB of RAM, that the post-processing step of the perf data would be OOM-killed by the kernel. When post-processing fails, the system is left in an awkward state that needs to be cleaned up, and it also may prevent the collection of remotely-registered data recordings.
We already set the perf recording frequency to 100Hz when we enable callgraphs. This patch sets the perf frequency to 100Hz by default to reduce the sampling size. In my experience this won't cause a loss of fidelity in trace data.