You’re likely running into a combination of resource exhaustion on EC2 and connector‑level performance limits, but the knowledge sources don’t give a precise recipe for S3 + multiline JSON, so I can’t be fully certain about the root cause. From what is documented: ### 1. EC2 instance overload Running many syncs or a heavy sync at once can easily overload a small EC2 instance. Airbyte jobs are resource‑hungry, and even for basic testing the team recommends at least a

t2.medium

, with larger instances for production. [[EC2 sizing](https://discuss.airbyte.io/t/2334)] If your instance is smaller than that, or you’re running many concurrent syncs, it’s expected that the machine becomes unresponsive. What you can tune: On self‑managed deployments you can limit concurrency and job resources: - Limit concurrent syncs (so you don’t “process too many files at once”):

yaml   worker:     maxSyncWorkers: 5        # or lower, depending on your instance     maxCheckWorkers: 5

(Helm v2 example.) [[Scaling Airbyte](https://docs.airbyte.com/platform/enterprise-setup/scaling-airbyte)] - Reduce per‑job CPU/memory so each sync is lighter:

yaml   global:     jobs:       resources:         limits:           cpu: 250m           memory: 500m         requests:           cpu: 75m           memory: 150m

[[Scaling Airbyte](https://docs.airbyte.com/platform/enterprise-setup/scaling-airbyte)] This prevents Airbyte from scheduling too many heavy pods at once and overwhelming the EC2 host. ### 2. Many connections / jobs = a lot of overhead Even when each dataset is small, a large number of connections or jobs creates significant overhead and can exhaust CPU/RAM. [[Airbyte resources](https://discuss.airbyte.io/t/995)] If you have many S3 connections or a single connection that spawns many streams, keep

MAX_SYNC_WORKERS

conservative so you don’t run dozens of jobs in parallel. Increasing

MAX_SYNC_WORKERS

increases parallelism and can speed things up, but only if your instance has enough headroom. [[MAX_SYNC_WORKERS](https://discuss.airbyte.io/t/1547); [Scaling Airbyte](https://docs.airbyte.com/platform/enterprise-setup/scaling-airbyte)] ### 3. Multiline JSON and S3 performance The docs and forum posts confirm that S3 JSON handling has been a pain point: - Historically, the S3 source did not support JSON/JSONL at all and needed a dedicated JSON parser to be added. [[S3 JSON enhancement](https://discuss.airbyte.io/t/1561); [S3 JSONL format](https://discuss.airbyte.io/t/539)] - There are also open performance concerns around S3‑based syncs being slow or stalling for larger volumes. [[Slow S3 sync](https://discuss.airbyte.io/t/3631)] However, the provided sources don’t explicitly mention a “multiline JSON is slow” warning, nor do they document specific tuning options for multiline JSON in the S3 source. So I can’t confirm from docs exactly how multiline JSON impacts performance in your version. Given what you’re seeing (long idle runs, EC2 overload, many files): - It’s plausible that parsing multiline JSON is CPU‑heavy, and - If the connection hasn’t run in a long time and is full refresh or has a loose file‑matching pattern, it may indeed be trying to process a very large backlog of files. ### 4. Practical steps within documented knobs Within what’s covered in the docs, you can: 1. Reduce concurrency and job size - Lower

MAX_SYNC_WORKERS

/

MAX_CHECK_WORKERS

. [[Scaling Airbyte](https://docs.airbyte.com/platform/enterprise-setup/scaling-airbyte)] - Reduce

jobs.resources.limits

in

values.yaml

so each sync uses less CPU/memory. [[Scaling Airbyte](https://docs.airbyte.com/platform/enterprise-setup/scaling-airbyte)] 2. Scale up the EC2 instance - Use at least a

t2.medium

for testing; larger for production or heavy S3 jobs. [[EC2 sizing](https://discuss.airbyte.io/t/2334)] 3. Check job logs for stalls / retries - Long jobs with “records extracted” but “0 loaded” often indicate the job stalled or failed after extraction, then retried, re‑reading the same data. The docs don’t show S3‑specific logs, but this pattern is consistent with other connectors. [[Connection timeline](https://docs.airbyte.com/platform/cloud/managing-airbyte-cloud/review-connection-timeline#syncs-refreshes-and-clears)] Because the knowledge sources don’t include your exact S3 + multiline JSON configuration, I can’t say more precisely why it’s overloading your instance. If you can, review the job logs for that S3 sync and consider: - Reducing the number of files matched (narrow the S3 path/prefix), and - Running a test on a subset of files to see if the job completes without overwhelming EC2.