TL;DR:
common Host size in CI's is 1-2G - for example Ubuntu Autopkgtests 1.5G
Some of them run Guests of 0.5-1G size in TCG mode (as they often can't rely on having KVM available).
The 1G TB buffer + 0.5G actual guest size + there is no dynamic downsizing on memory pressure (never was) makes these systems go OOM-Killing the qemu process.
Let us try to tune back the default size a bit again while the upstream discussion is ongoing.
As I reported on /lists. gnu.org/ archive/ html/qemu- devel/2020- 07/msg05001. html
https:/
TL;DR:
common Host size in CI's is 1-2G - for example Ubuntu Autopkgtests 1.5G
Some of them run Guests of 0.5-1G size in TCG mode (as they often can't rely on having KVM available).
Due to https:/ /git.qemu. org/?p= qemu.git; a=commit; h=600e17b261555 c56a048781b8dd5 ba3985650013 in qmeu 5.0 the TB default size bumped from 32M to 1G.
The 1G TB buffer + 0.5G actual guest size + there is no dynamic downsizing on memory pressure (never was) makes these systems go OOM-Killing the qemu process.
Let us try to tune back the default size a bit again while the upstream discussion is ongoing.