Maggie Nolan Edmonds
Maggie Nolan Edmonds
The profile.proto data format does not have a specific way to note which shared object a sample is associated with. When using pprof on perf.data files, it should be possible...
Right now, only values associated with the "bytes" tag will ever be displayed in any call graph views. This is because of the following section of code: https://github.com/google/pprof/blob/master/internal/report/report.go#L244-L253 We should...
The profile.proto data format does not have a specific way to note which process a sample came from. However, perf.data files tend to include samples from multiple processes. When using...
Right now, `0B:256B` is interpreted at a range for filtering tags, but `0:256B` is not. I think we should either infer that the units are the same for both parts...
A customer indicated that we should review this document. They tried to enable profiler when running on GKE, but weren't able to because profiler required that the service name be...
The request in createProfile() can hang for up to an hour. So, when a user's code finishes running, it could be up to an hour before profiler exits (and node.js...
Related to #19, but focusing on heap profile.
This builds on #204. By lazily loading source maps, we should be able to start reading source maps in the node_modules directory.
ServerBackoffResponse (in pr #75) corresponds to a server response with RetryInfo in error_details.proto (here https://github.com/googleapis/googleapis/blob/master/google/rpc/error_details.proto). We should use (or at least investigate using) error_details.proto.
@google-cloud/cloud-debug-nodejs uses the cluster name for the service (https://github.com/GoogleCloudPlatform/cloud-debug-nodejs/blob/a828ea62a434e3c92eda46e39921531ba30e68cf/src/agent/debuglet.ts#L347). Profiler should do the same.