[metrics] Force export census metrics on worker death #28547
Merged
Conversation
This file contains 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
Signed-off-by: Eric Liang <[email protected]>
Signed-off-by: Eric Liang <[email protected]>
Signed-off-by: Eric Liang <[email protected]>
Signed-off-by: Eric Liang <[email protected]>
Signed-off-by: Eric Liang <[email protected]>
Signed-off-by: Eric Liang <[email protected]>
Signed-off-by: Eric Liang <[email protected]>
Signed-off-by: Eric Liang <[email protected]>
Signed-off-by: Eric Liang <[email protected]>
Signed-off-by: Eric Liang <[email protected]>
Signed-off-by: Eric Liang <[email protected]>
Signed-off-by: Eric Liang <[email protected]>
Signed-off-by: Eric Liang <[email protected]>
Signed-off-by: Eric Liang <[email protected]>
Signed-off-by: Eric Liang <[email protected]>
ericl
commented
Sep 15, 2022
| @@ -460,7 +460,7 @@ RAY_CONFIG(int64_t, idle_worker_killing_time_threshold_ms, 1000) | |||
| RAY_CONFIG(int64_t, num_workers_soft_limit, -1) | |||
|
|
|||
| // The interval where metrics are exported in milliseconds. | |||
| RAY_CONFIG(uint64_t, metrics_report_interval_ms, 10000) | |||
| RAY_CONFIG(uint64_t, metrics_report_interval_ms, 5000) | |||
There was a problem hiding this comment.
Unrelatedly, it seems 5 seconds is probably safe for now.
rkooo567
approved these changes
Sep 16, 2022
| @@ -590,6 +590,8 @@ void CoreWorker::Disconnect( | |||
| const rpc::WorkerExitType &exit_type, | |||
| const std::string &exit_detail, | |||
| const std::shared_ptr<LocalMemoryBuffer> &creation_task_exception_pb_bytes) { | |||
| // Force stats export before exiting the worker. | |||
| opencensus::stats::StatsExporter::ExportNow(); | |||
There was a problem hiding this comment.
Is this method thread safe?
Also, technically this can lose data because ExportNow() is not a blocking call (so if metrics_agent_io_service stops before we send a RPC, it can lose data). I guess the probably is low, and it might be okay.
There was a problem hiding this comment.
Yes, I checked here and OCL acquires a mutex internally.
Regarding the RPC safety, I think at least the RPC initiation is a blocking call. I'm not sure if we wait for a reply though.
There was a problem hiding this comment.
Ah, I see. If the initiation is a blocking call, I think it is pretty safe although we are not waiting for a reply...
Signed-off-by: Eric Liang <[email protected]>
Signed-off-by: Eric Liang <[email protected]>
stephanie-wang
added a commit
that referenced
this pull request
Sep 19, 2022
This reverts commit 54136e8.
stephanie-wang
added a commit
that referenced
this pull request
Sep 19, 2022
ericl
added a commit
to ericl/ray
that referenced
this pull request
Sep 19, 2022
PaulFenton
pushed a commit
to PaulFenton/ray
that referenced
this pull request
Sep 19, 2022
) Signed-off-by: PaulFenton <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Why are these changes needed?
Census metrics are periodically pushed from individual workers to per-node metrics agent processes for aggregation. However, this can result in missing the latest metrics from workers (e.g., last few tasks that finished). Force an export during core worker shutdown to fix this.
This involves patch OCL to allow access to the ExportNow() API.