samueldr has quit [Ping timeout: 260 seconds]
samueldr_ has joined #nixos-systemd
hmpffff has joined #nixos-systemd
hmpffff has quit [Ping timeout: 256 seconds]
hmpffff has joined #nixos-systemd
emily has quit [Quit: killed]
colemickens has quit [Quit: killed]
hmpffff[m] has quit [Quit: killed]
JJJollyjim has quit [Quit: killed]
Ox4A6F has quit [Quit: killed]
jtojnar has quit [Quit: killed]
worldofpeace has quit [Quit: killed]
ma27[m] has quit [Quit: killed]
balsoft has quit [Quit: killed]
danielrf[m] has quit [Quit: killed]
aanderse has quit [Quit: killed]
Ox4A6F has joined #nixos-systemd
hmpffff has quit [Ping timeout: 272 seconds]
hmpffff has joined #nixos-systemd
colemickens has joined #nixos-systemd
aanderse has joined #nixos-systemd
emily has joined #nixos-systemd
hmpffff[m] has joined #nixos-systemd
JJJollyjim has joined #nixos-systemd
worldofpeace has joined #nixos-systemd
danielrf[m] has joined #nixos-systemd
ma27[m] has joined #nixos-systemd
balsoft has joined #nixos-systemd
jtojnar has joined #nixos-systemd
hmpffff has quit [Ping timeout: 260 seconds]
hmpffff has joined #nixos-systemd
hmpffff has quit [Ping timeout: 240 seconds]
hmpffff has joined #nixos-systemd
<gchristensen> I'm seeing the io controller enabled in /sys/fs/cgroup/cgroup.controllers, /sys/fs/cgroup/system.slice/cgroup.controllers, and /sys/fs/cgroup/system.slice/nix-daemon.service/cgroup.controllers with IOAccounting enabled on system.slice and nix-daemon.service, but /sys/fs/cgroup/system.slice/nix-daemon.service/io.stat is totally empty. And actually, almost all of my io.stat files are empty,
<gchristensen> and systemd-cgtop doesn't show any data for inputs and outputs. here are all the cgroups with present io.stat data, and all the cgroups without: https://gsc.io/content-addressed/9e6374452d113c3c0ab798bd70671f4b3e631a6e803620346d61b42e5e3bd70a -- any ideas what isgonig on? systemd 246, cgroupv2, nixos 20.09beta1346, Linux 5.8.16
qyliss has quit [Remote host closed the connection]
qyliss has joined #nixos-systemd
aminechikhaoui has quit [Ping timeout: 240 seconds]
aanderse has quit [*.net *.split]
colemickens has quit [*.net *.split]
aanderse has joined #nixos-systemd
colemickens has joined #nixos-systemd
<Mic92> arianvp: I wonder how many application will still break/ be incompatible with varlink and its nss modules.
aminechikhaoui has joined #nixos-systemd
samueldr_ is now known as samueldr
samueldr has joined #nixos-systemd
samueldr has quit [Changing host]
hmpffff has quit [Ping timeout: 272 seconds]
hmpffff has joined #nixos-systemd
<flokli> https://www.freedesktop.org/software/systemd/man/systemd.resource-control.html#IOAccounting= says "if the unified control group hierarchy is used on the system"
<flokli> we still set it to hybrid by default, because we don't want to break docker, at least not until some of the open PRs with /var/lib/containers have been fixed
<gchristensen> I use the unified control group hierarchy, not hybrid
<flokli> that being said, we might want to switch that default for 20.09. See https://medium.com/nttlabs/cgroup-v2-596d035be4d7 and https://github.com/systemd/systemd/blob/master/docs/CGROUP_DELEGATION.md#
<flokli> hm, then I don't know.
<gchristensen> (if you were responding to my question. if you're talking about something else, I'll stay out :P)
<flokli> I spoke about your question
<flokli> all the io accounting highly depends on cgroups. so which of the three possible hierarchies your using matters ;-)
<gchristensen> my kernel cmdline includes cgroup_no_v1=all systemd.unified_cgroup_hierarchy=yes
<gchristensen> maybe I'm missing something else? (though /sys/fs/cgroup is indeed, cgroup2fs)
<flokli> I think so far I've only seen Network traffic (IPAccounting=) to show something, but not disk i/o.
<gchristensen> yeah, I generally see that working -- evidence in the service stop messages, summarizing raffic
<gchristensen> I guess I shouldn't ask then if you have v2 enabled, and if you see IOAccounting data :D
<flokli> I use the defaults, which currently is unified.
<gchristensen> hybrid
<flokli> ergh, yes
<flokli> sorry
<gchristensen> okay, no worries, cool
<flokli> I use hybrid, which is the NixOS default currently.
<flokli> this should also be in systemctl status of a unit since https://github.com/systemd/systemd/pull/12074
<{^_^}> systemd/systemd#12074 (by poettering, 1 year ago, merged): expose IO stats on the bus and in "systemctl status" and "systemd-run --wait"
<gchristensen> is this question worth retrying in #systemd, or is there something I should be reading first? (I tried to be thorough {reading the resource control man page, and the linux kernel docs on the cgroupsv2 documentation, especially on the io controller} before coming here, but if you have further pointers...)
<flokli> I haven't seen it working so far, but didn't bother enough to check further. If you want to drill down into it, I'd be more than interested to see why it's currently not working, and how we can fix it.
<gchristensen> cool, thanks
NinjaTrappeur has quit [Ping timeout: 260 seconds]
hmpffff has quit [Ping timeout: 272 seconds]
hmpffff has joined #nixos-systemd
asymmetric has quit [Ping timeout: 246 seconds]
asymmetric has joined #nixos-systemd
NinjaTrappeur has joined #nixos-systemd
aminechikhaoui has quit [*.net *.split]
aminechikhaoui has joined #nixos-systemd
globin has quit [Ping timeout: 264 seconds]
globin has joined #nixos-systemd
hmpffff has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]