[BUG] Fatal OOM: JavaScript heap out of memory during Admin relational data fetches

[LukaGiorgadze]

Summary

The server process is consistently crashing with a FATAL ERROR: Reached heap limit Allocation failed - JavaScript heap out of memory. This appears to be triggered when the Admin Dashboard attempts to fetch seller and its orders/products with deep relational expansions.

FATAL ERROR: Reached heap limit Allocation failed - JavaScript heap out of memory
[39:0x9cea8c0] 2329014 ms: Scavenge 813.1 (832.5) -> 812.2 (836.8) MB, 136.42 / 0.00 ms
<--- JS stacktrace --->
1: 0xb71e90 node::OOMErrorHandler
2: 0xeaabdc v8::Utils::ReportOOMFailure
...

Context & Observed Behavior

Prior to the crash, several high-load requests were observed targeting the admin API. Specifically, requests using wildcards (*) for field expansion seem to push the memory usage past the ~832MB limit:

GET /admin/stores
GET /admin/stock-locations?q=&limit=3&fields=id%2Cname
GET /admin/sellers/.../customer-groups?fields=...,*customers
GET /admin/sellers/.../orders?fields=...,*customer
GET /admin/orders?fields=id,display_id,email

Environment: Medusa Cloud [Production]

Possible Root Causes

• Inefficient Query Expansion: The use of *customers and *orders in the admin dashboard fetches may be attempting to load thousands of records into the Node.js heap simultaneously without sufficient pagination or selective field loading.
• Heap Memory Constraints: The process is hitting a hard limit at ~800MB. MercurJS/Medusa v2 may require a higher max-old-space-size when dealing with large multi-vendor datasets.
• Relation Hydration: The way MercurJS hydrates relations for sellers might be creating large object trees that the Garbage Collector cannot clean up fast enough.

Steps to Reproduce

1. Navigate to the Admin Dashboard
2. Access a Seller with a significant number of products and orders.
3. Observe memory usage spike until the process terminates.

[olo-rigby]

Thanks for this one. We are looking into it.