Hi Michael, Michael Ivanov <ivans@isle.spb.ru> writes:
Greetings,
I am running oracle using podman volume for data (it is not for production, of course, development only). So far I did not notice anything noticeably wrong about it, but today I observed, that adump directory has a lot of files and decided to clean it up (just rm -rf adump). As of time of writhing the command runs for about half an hour and did not yet complete. Ten minutes ago I checked number of remaining files and it was about 350000. The removal goes at a rate on about 1000 files per minute or so. On hosting system fuse-overlayfs process takes 99% cpu.
that is really slow, even for fuse-overlayfs. Could you share more information about your system?
My system is rocky linux (former centos) 8.5. I installed podman from standard rocky linux repo, no external repos were used. Podman version is3.2.3, crun 0.20.1 is used fpr container runtime. Underlying fs for container storage is ext4. I did not do any special setup/configuration at all, apart from creating an lvm for co=ntainer storage and mounting it under /home/xxxx/.local/share/containers (podman runs rootless).
Is this behaviour normal?
I suggest to switch to native overlay for rootless, that will help with the issue you've seen. You need to fully destroy your storage first "podman system reset -f", and next time you run Podman it will detect the native overlay support and use it instead of fuse-overlayfs. Regards, Giuseppe