Dnf-makecache.timer is a major liability

Rocky Linux Tech Help
1

Posting this here to exchange notes before I go to wherever dnf lives.

For utterly baffling reasons, the version of dnf on Rocky Linux 9.x (current one is 4.14) needs literally gigabytes of memory for dnf makecache --timer. Several of my VMs run at capacity, and the spike of memory caused by this job leaves the box swapping or frozen.

I’ve been happily disabling this unit across all RHEL 9 boxes for maybe a year now but it stung me again on a new box a few days ago (I forgot to disable it) and now I’m wondering if anyone out there is even aware of this issue.

2

Upstream (Fedora) maintainers appear to be aware:

https://bugzilla.redhat.com/show_bug.cgi?id=1907030

1 Like
3

But this is only on vms that are set up with low memory.

It’s good they are trying to improve how DNF works, but once they fix this, some other process will end up being killed by low memory.

4

Note: the last time this happened I had 1.5GB of headroom. That’s… a lot of memory for a package manager.

5

I routinely disable it, systemctl disable dnf-makecache.timer - which is easy if you utilize VM templates in vmware. The purpose for which it’s enabled by default eludes my comprehension…

6

This is likely related to a problem I saw; RL9 dnf OOMs with EPEL but only on one machine!

1 Like
7

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.