Chrome may use wrong memory-available numbers when running in a cgroup |
||
Issue descriptionChrome Version: (copy from chrome://version) Version 66.0.3359.117 (Official Build) (64-bit) OS: (e.g. Win10, MacOS 10.12, etc...) Linux (glinux) What steps will reproduce the problem? (1) Enter the following script as "mchrome": #!/bin/sh sudo mkdir -p /sys/fs/cgroup/memory/chrome sudo chmod -R a=rwX /sys/fs/cgroup/memory/chrome echo 12G > /sys/fs/cgroup/memory/chrome/memory.limit_in_bytes echo 0 > /sys/fs/cgroup/memory/chrome/memory.swappiness exec cgexec -g memory:chrome google-chrome (2) Run it. (3) Stress Chrome - open lots of tabs and documents. What is the expected result? Chrome should run normally, using no more than 12G of RAM What happens instead? Chrome crashes more frequently than usual, and frequently goes into CPU-consuming loops that clog up the system. Example crash (I think): f9ac030f1d9cc5a6 Suspected cause: Chrome uses sysctl calls and /proc file reading (see process_metrics_linux.cc) that don't account for containers, and thus delays garbage collection until after the process has already run out of memory.
,
May 22 2018
,
Jun 9 2018
I have recently gone back to using this configuration, and crashes are indeed more frequent than otherwise. Crash Report ID 468cea22fa696519 (browser) Crash Report ID 64688497d812ef4a (renderer) |
||
►
Sign in to add a comment |
||
Comment 1 by hta@chromium.org
, Apr 24 2018