Re, > No I've tried some things and found out that these processes caused > needrestart to believe that libvirtd needs to be restarted: > # ps ax | grep libvirt > 10677 ? Ssl 0:00 /usr/sbin/libvirtd > 11264 pts/0 S+ 0:00 grep --color=auto libvirt > 31720 ? S 0:00 /usr/sbin/dnsmasq > --conf-file=/var/lib/libvirt/dnsmasq/main.conf > --dhcp-script=/usr/lib/libvirt/libvirt_leaseshelper 31721 ? > S 0:00 /usr/sbin/dnsmasq > --conf-file=/var/lib/libvirt/dnsmasq/main.conf > --dhcp-script=/usr/lib/libvirt/libvirt_leaseshelper 31779 ? > S 0:00 /usr/sbin/dnsmasq > --conf-file=/var/lib/libvirt/dnsmasq/insecure.conf > --dhcp-script=/usr/lib/libvirt/libvirt_leaseshelper 31780 ? > S 0:00 /usr/sbin/dnsmasq > --conf-file=/var/lib/libvirt/dnsmasq/insecure.conf > --dhcp-script=/usr/lib/libvirt/libvirt_leaseshelper i.e. after > killing them (without restarting libvirtd another time) made the > needrestart suggestion to restart libvirtd go away.
the dnsmasq processes are using an outdated binary file which is detected by needrestart. Their parent process is the libvirtd which is used to get the systemd service name / find a sysv init script. > So I guess the problem is actually that the libvirtd init-script / > unit files don't handle the restart "correctly"[0] and let their > spawned processes alive. ACK > Now I don't know need restart detects which init-script / unit file it > needs to restart when a given process uses outdated libs. > With systemd only it would be clear, cause every service runs in it's > cgroup, but with sysvinit? > Anyway if you agree this is the cause for this (not-so-)false > positive, then we can retitle and reassign this to libvirtd. Yes, I'm able to confirm your results from the verbose output of needrestart. > Of cause one issue remains, actually two, but with respect to > needrestart they both lead to the same question and I'll open a > separate bug for it. > Is it really the task/duty of libvirtd's initscript/unit file to > restart dnsmasq, which it could also just consider to be another > "external" service. > So libvirtd maintainers/upstream could argue,... "we just start > dnsmasq for some arbitrary virtual networks which, even though set up > by our tools and used with our VMs, are still generic networks, which > we just start once but not maintain thereafter". > So from a needrestart POV the question is: shouldn't it tell the user > somehow (in non-verbose mode) which processes (in terms of PID) it > think should be restarted? Needrestart misses a detailed report like checkrestart does - see also #762896. To be more verbose does not help with this bug ;-) Cheers, Thomas -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
