EPS image upload timeout/failure 14-APR-2026
I'm running into an EPS photo ingestion issue today. I'm getting consistent timeouts from the EPS service. I'll send over a handful of files with no issue, then the next file will result in a timeout.
Here's the request:
{
 "call": "createImageFromFile",
 "local_path": "/srv/ai-platform/data/photo_pipeline/approved/SKU_3FB44C10/Photo Apr 12 2026, 4 48 40 PM.jpg",
 "filename": "Photo Apr 12 2026, 4 48 40 PM.jpg",
 "bytes": 2303274
}
Here's the reply:
Here's the reply:
{
 "http_status": null,
 "duration_sec": 91.561,
 "exception_type": "SSLEOFError",
 "exception_text": "<urlopen error EOF occurred in violation of protocol (_ssl.c:2406)>",
 "error_summary": "remote TLS endpoint closed the connection unexpectedly during upload (SSLEOFError)"
}
The error isn't the same every time. I will occasionally receive a "HTTP 503 Internal Server Error - Write". I've tried multiple items with the same result.
If the EPS ingestion engine was broken like this, I'd think that it would be a BIG deal for anyone using the API.
Anyone else seeing this? I sure hope it's transitory and clears soon. 🙂
-Jason
The error isn't the same every time. I will occasionally receive a "HTTP 503 Internal Server Error - Write". I've tried multiple items with the same result.
If the EPS ingestion engine was broken like this, I'd think that it would be a BIG deal for anyone using the API.
Anyone else seeing this? I sure hope it's transitory and clears soon. 🙂
-Jason
jwhite7137
Posted 4 months ago·Last reply 4 months ago
2 comments
jwhite7137
OP4 months agoFYI - it turns out that this issue was caused by a problem at my ISP. Specifically, the ISP broke Path MTU Discovery. I had to dial back the MTU frame size on my router to them to 1460 bytes. This is a hack, but gets around the problem until they decide to fix it.
After dialing back the MTU, I've been able to upload images to the EPS ingestion service and list items via the API.
This was a network engineering issue, not an issue with ebay services.
jwhite7137
OP4 months agoHere's a sanitized HTTP 500 response after testing the same workflow again: