Post At8pE1JBdlVc0vD5FY by feld@friedcheese.us
(DIR) More posts by feld@friedcheese.us
(DIR) Post #At45GjgKGAo1grYBRA by Tubsta@soc.feditime.com
2025-04-13T22:49:19.840491Z
1 likes, 0 repeats
Memory management on #Linux is a bit borked. I tried to ingest 4GB of records from a file into a list (array) with #Python. Instead of the machine releasing RAM, it decided just to use swap and before I knew it, swapped was consumed and the Python process fell over.I'll move the workload over to one of my #FreeBSD lab hosts tonight and re-run the script.
(DIR) Post #At8pDztctJT7dMF9lo by ianthetechie@fosstodon.org
2025-04-14T00:12:22Z
0 likes, 0 repeats
@Tubsta curious to hear a postmortem report on that + how it runs on FreeBSD. Can’t say I’ve ever hit that issue before. (While I really wish Linux would adopt jemalloc, I rather doubt that’s to blame..)
(DIR) Post #At8pE1JBdlVc0vD5FY by feld@friedcheese.us
2025-04-14T00:23:25.530245Z
0 likes, 0 repeats
@ianthetechie @Tubsta there may be a new allocator in FreeBSD in the not too distant future toohttps://github.com/microsoft/snmalloc
(DIR) Post #At8pE2EcCFRit24wlM by feld@friedcheese.us
2025-04-14T00:28:06.292283Z
0 likes, 0 repeats
@ianthetechie @Tubsta the CFT for it was announced herehttps://lists.freebsd.org/archives/freebsd-hackers/2023-February/001808.html
(DIR) Post #At8pE34N5oqXTYIHR2 by ianthetechie@fosstodon.org
2025-04-14T00:40:40Z
0 likes, 0 repeats
@feld @Tubsta ooh wow! I didn’t hear about this! Thanks for sharing!
(DIR) Post #At8pE4AmzRZetkIv4q by Tubsta@soc.feditime.com
2025-04-15T22:23:04.288625Z
1 likes, 0 repeats
@ianthetechie @feld I can confirm that #Python on #FreeBSD behaves as one would expect. It consumes all RAM (with #ZFS releasing ARC as expected) and then dips into swap. As soon as Python releases memory after the ingestion routine, the swap is purged to near zero and the RAM then becomes available (and used) by the system. Far more predictable and reliable.If you have big, vertical workloads, FreeBSD is where it is at.