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
Under high concurrency testing, latency got slower, timeouts started to happen, and we get intermittent " libgomp: Thread creation failed: Resource temporarily unavailable".
Not sure how best to characterize the "concurrency", but on CPU I'd say we test w/ 100 concurrent requests and see latency degrade and at 500 we start getting errors.
This is a regression that showed up in 0.25.6.
NOTE! It looks like configuring server_thread_pool_size: 5 fixes it. In 0.25.6 that default was changed from 5 to 100 (related to HTTP issues). We are only using gRPC.
Platform
Please provide details about the environment you are using, including the following:
Interpreter version:
Library version: Comparing 0.25.x versions
Sample Code
Please include a minimal sample of the code that will (if possible) reproduce the bug in isolation
Expected behavior
In my case, it seems like the default should be the better performing 5.
Ideally limits would prevent the resource error from ever happening (but don't know how possible that is to fix).
Observed behavior
What you see happening (error messages, stack traces, etc...)
Additional context
Add any other context about the problem here.
The text was updated successfully, but these errors were encountered:
Describe the bug
Under high concurrency testing, latency got slower, timeouts started to happen, and we get intermittent " libgomp: Thread creation failed: Resource temporarily unavailable".
Not sure how best to characterize the "concurrency", but on CPU I'd say we test w/ 100 concurrent requests and see latency degrade and at 500 we start getting errors.
This is a regression that showed up in 0.25.6.
Platform
Please provide details about the environment you are using, including the following:
Sample Code
Please include a minimal sample of the code that will (if possible) reproduce the bug in isolation
Expected behavior
In my case, it seems like the default should be the better performing 5.
Ideally limits would prevent the resource error from ever happening (but don't know how possible that is to fix).
Observed behavior
What you see happening (error messages, stack traces, etc...)
Additional context
Add any other context about the problem here.
The text was updated successfully, but these errors were encountered: