Open remingtonc opened 5 years ago
Seeing this hit ~60 GB of RAM usage. It appears that ArangoDB is computing the entire query, and then streaming the results in the batch size. My impression was that computation and results would both be on the fly and per batch size - need to investigate.
We could move the search entry into the YANG parsing section...
Tracking in etl-search-stream
.
The dbms service is getting OOMurdered during ETL query. #36
Likely due to flattening in query. Should move to streaming results. https://github.com/cisco-ie/tdm/blob/05eb9c91ddefc8dfa21c919bec880a4d3a2d9fcc/etl/src/search.py#L39