Launching VM over RBD backend crashes the compute node
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Fuel for OpenStack |
Invalid
|
Critical
|
Unassigned |
Bug Description
Fuel 4.0 ISO #214, non-HA, Ubuntu, controller, 2x compute, all Ceph options enabled (including Ceph RBD backend for ephemeral volumes).
After creating a RAW CirrOS image and launching a VM from it using following commands:
qemu-img convert -O raw /usr/share/
glance image-create --name cirros-raw --is-public yes --container-format bare --disk-format raw < /usr/share/
nova boot --flavor 1 --image cirros-raw vm0
the compute node that tries to launch the image locks up, last messages in the logs are:
==> /var/log/daemon.log <==
<30>Dec 30 18:23:42 node-7 ntpd[17142]: Listen normally on 17 br100 10.108.2.5 UDP 123
<30>Dec 30 18:23:42 node-7 ntpd[17142]: Listen normally on 18 br100 fe80::bc84:
<30>Dec 30 18:23:42 node-7 ntpd[17142]: peers refreshed
<31>Dec 30 18:23:42 node-7 ntpd[17142]: new interface(s) found: waking up resolver
<30>Dec 30 18:23:43 node-7 dnsmasq[8559]: started, version 2.59 cachesize 150
<30>Dec 30 18:23:43 node-7 dnsmasq[8559]: compile time options: IPv6 GNU-getopt DBus i18n DHCP TFTP conntrack IDN
<30>Dec 30 18:23:43 node-7 dnsmasq-dhcp[8559]: DHCP, static leases only on 10.108.2.2, lease time 2m
<30>Dec 30 18:23:43 node-7 dnsmasq[8559]: reading /etc/resolv.conf
<30>Dec 30 18:23:43 node-7 dnsmasq[8559]: using nameserver 10.108.0.2#53
<27>Dec 30 18:23:43 node-7 dnsmasq[8559]: failed to load names from /var/lib/
<30>Dec 30 18:23:43 node-7 dnsmasq-dhcp[8559]: read /var/lib/
==> /var/log/kern.log <==
<5>Dec 30 18:23:44 node-7 kernel: [253594.093269] type=1400 audit(138842782
The problem does not occur on CentOS.