stgraber indicates this is causing issues in lxc containers:
11:14 < stgraber> bdmurray: your unattended-upgrade SRU is making all my containers crash once an hour ;)
11:14 < bdmurray> stgraber: hunh?
11:15 < stgraber> you can't set negative nice values in an unprivileged LXC container even if you're root inside it
11:15 < stgraber> so now every time unattended-upgrade runs (in my case every hour), it crashes on that os.nice call
11:18 < stgraber> what we've done in other tools (like systemd) is make this non-fatel if the error is EACCES (typical for unprivileged containers) or EPERM (typical for apparmor preventing it)
11:19 < bdmurray> stgraber: but then wouldn't bug 1422345 still happen?
11:22 < stgraber> bdmurray: sure, the bug would still happen in this case
11:23 < stgraber> bdmurray: root in a userns has the same rights as a regular user with regard to process priorities
11:23 < stgraber> bdmurray: and as a regular user, you can increase the niceness but never decrease it
11:24 < bdmurray> would it better to first try decreasing and only increase it if decreasing works?
11:25 < stgraber> yep, that'd be a good trick
11:25 < stgraber> so -1, if that works then 20, otherwise keep is it is
stgraber indicates this is causing issues in lxc containers:
11:14 < stgraber> bdmurray: your unattended-upgrade SRU
is making all my containers crash
once an hour ;)
an unprivileged LXC container even if
you' re root inside it
runs (in my case every hour), it
crashes on that os.nice call
systemd) is make this non-fatel if
the error is EACCES (typical for
unprivilege d containers) or EPERM
(typical for apparmor preventing it)
1422345 still happen?
happen in this case
same rights as a regular user with
regard to process priorities
can increase the niceness but never
decrease it
decreasing and only increase it if
decreasing works?
otherwise keep is it is
11:14 < bdmurray> stgraber: hunh?
11:15 < stgraber> you can't set negative nice values in
11:15 < stgraber> so now every time unattended-upgrade
11:18 < stgraber> what we've done in other tools (like
11:19 < bdmurray> stgraber: but then wouldn't bug
11:22 < stgraber> bdmurray: sure, the bug would still
11:23 < stgraber> bdmurray: root in a userns has the
11:23 < stgraber> bdmurray: and as a regular user, you
11:24 < bdmurray> would it better to first try
11:25 < stgraber> yep, that'd be a good trick
11:25 < stgraber> so -1, if that works then 20,