Goal Determine the default HTTP connection pool size used by the major Sentry SDKs and the documented behavior when that pool is exhausted. Constraints Pool configuration is SDK‑specific (e.g., OkHttp for Java, urllib3 for Python, Netty for JavaScript) and not centrally managed. SDK documentation lists tuning parameters but does not expose built‑in metrics for pool saturation. Observed behavior varies: some transports queue indefinitely, others drop events after an undocumented timeout. Open questions What is the default connection pool size for the sentry‑python SDK's urllib3 transport? Does the sentry‑java SDK's OkHttp transport queue requests indefinitely or discard events after a specific timeout when the pool is full? Are there any version‑dependent timeout values that control event discard on pool saturation across SDKs?

Sentry SDK HTTP connection pool limits and exhaustion behavior

Sentry SDK HTTP connection pool default limits and exhaustion behavior

0 reputation · 09 Apr 2024, 19:54 UTC

80.5K views0

Goal

Determine the default HTTP connection pool size used by the major Sentry SDKs and the documented behavior when that pool is exhausted.

Constraints

Pool configuration is SDK‑specific (e.g., OkHttp for Java, urllib3 for Python, Netty for JavaScript) and not centrally managed.
SDK documentation lists tuning parameters but does not expose built‑in metrics for pool saturation.
Observed behavior varies: some transports queue indefinitely, others drop events after an undocumented timeout.

Open questions

What is the default connection pool size for the sentry‑python SDK's urllib3 transport?
Does the sentry‑java SDK's OkHttp transport queue requests indefinitely or discard events after a specific timeout when the pool is full?
Are there any version‑dependent timeout values that control event discard on pool saturation across SDKs?

Keep exploring this topic

What determines HikariCP pool sizing in JHipster-generated Spring Boot applications?

JHipster generates Spring Boot applications using HikariCP as the default connection pool implementation. The default maximum pool size is calculated as (availableProcessors * 2) + availableDiskParallelism, capped at 10, which may be insufficient for production workloads.

ActiveRecord connection pool exhaustion intermittently under multithreaded requests

Intermittent ActiveRecord connection pool exhaustion under multithreaded requests, with uncertainty about version differences and configuration impact.

Connection pool exhausted in persistent library

The persistent library throws a ConnectionPoolExhausted exception when all pool connections are busy. The community is undecided on whether to default to blocking or fast-failing.

Exceeding Allocated Limits in Apex Connection Pool

Investigating the precise mechanism for connection pool exhaustion in Apex, specifically whether it relates to unclosed resources or defined runtime configuration limits.

A useful clarification: in sentry-java, OkHttp dispatcher queuing is separate from Sentry's internal event queue. OkHttp may queue calls when dispatcher limits are reached, but Sentry Java can drop events from its bounded queue before the transport experiences pool pressure, so pool metrics alone won't explain missing events.

For sentry-python, urllib3's default block=False generally opens a new connection when a pool is full rather than blocking, though this depends on the SDK's PoolManager arguments and urllib3 version. Verify by inspecting the transport constructor in your SDK release and load-testing against a local endpoint to observe whether calls block, queue, or fail.

Sentry SDK HTTP connection pool default limits and exhaustion behavior

Goal

Constraints

Open questions

0 answers

Your perspective belongs here.

1 question comment