Instances stuck in spawning state due to Nova logging deadlock
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Fuel for OpenStack |
Triaged
|
High
|
Roman Vyalov |
Bug Description
Steps to reproduce:
1. Install Fuel iso on kvm.
2. Create new cluster with the following configuration:
- centos multinode;
- nova-network flat;
- enable ceph for volumes;
- enable ceph for ephemeral storage;
3. Add the following nodes:
- controller;
- compute + ceph;
- compute + ceph;
4. Deploy cluster - wait until deployment is successful;
5. Check all services are running;
6. Go to Horizon;
7. Create new instance
Expected result:
Instance is created and is in active state;
Actual result:
Instance goes to spawning state which never ends;
The longest time it was in spawning state was 20 hours and spawning was not interrupted by timeout. Instance got stuck into this state right after getting network on compute.
In fact instance was not created at all - it couldn`t receive network. virsh list --all on compute-node where instance should`ve been located shows no vms at all. virsh net-list --all shows only default network which is inactive.
The following errors found in libvirt logs on compute:
2014-02-11 13:51:34.828+0000: 16498: info : libvirt version: 0.10.2, package: 19.el6 (Unknown, 2013-10-
2014-02-11 13:51:34.828+0000: 16498: error : virNetDevBridge
2014-02-11 13:51:34.898+0000: 16498: error : virNWFilterSnoo
2014-02-11 13:51:34.963+0000: 16498: warning : virSecurityMana
2014-02-11 14:06:32.882+0000: 16486: error : virNetSocketRea
Reproduced on iso 108, iso fuel-stop_
This bug is actually a duplicate of https:/ /bugs.launchpad .net/nova/ +bug/1276268 . Ticket for cherry-pick created for OSCI team https:/ /mirantis. jira.com/ browse/ OSCI-1071