I ran a script this afternoon that died mys­te­ri­ous­ly without any output. It was using SQLAlchemy to query all the rows from a large table so that they could be trans­formed into JSON Lines to be loaded into Elas­tic­search. When I reran my script, I noticed this time that something had printed Killed at the very end.

A little research convinced me that the OOM Killer was the likely assassin. I looked in /var/log/kern.log and I found that my process had used up almost all of the 8GB on this system before being killed.

The query had to be the problem. A little more research led me to augment my query with yield_per, which batched the results, instead of fetching everything into memory at once. There are various caveats to using yield_per; none applied here.