Repeated boot of ironic instances on added conductor failed
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Mirantis OpenStack |
Confirmed
|
Medium
|
MOS Ironic |
Bug Description
Steps:
1. Prepared cluster configuration:
1 Controller
1 Ironic
1 compute
1 not allocated node
Compute - QEMU
Network - Neutron with VLAN segmentation
Storage Backends - Cinder LVM over iSCSI for volumes
2. Deployed it to the operational state.
3. Then booted Ironic instance VM according to happy path Configuration Instruction.
4. Shutdown ironic instance with "nova delete" command.
5. Add 1 more Ironic Conductor node to Cluster.
6. Deploy changes.
7. Boot Ironic instance VM. Shutdown it and boot more several times.
Actual result:
Only twice VM booted successfully.
Inspection of ironic-conductor logs discovered that after redeployment only 2nd conductor was used.
Here is snippet of 1st Conductor log: http://
Note that last updates from 2015-12-14 22:36:38.
UPD: looks that it is expected when only one conductor is used while load is very low.
Conductor Log snippet from 2nd Conductor node:
http://
Please note there two errors occurred:
AttributeError: 'ConductorManager' object has no attribute '_keepalive_e
vt'
and
Cannot execute SSH cmd LC_ALL=C /usr/bin/virsh --connect qemu:/
//system destroy kromanenko-
VM on virtual manager unexpectedly gone offline.
When I switched ON it back - it booted Ubuntu that was deployed on last successful nova boot.
VERSION:
feature_groups:
- mirantis
production: "docker"
release: "8.0"
openstack_
api: "1.0"
build_number: "275"
build_id: "275"
description: | updated |
Changed in mos: | |
status: | New → Confirmed |
tags: | added: mos-ironic |
description: | updated |
description: | updated |
description: | updated |
"AttributeError: 'ConductorManager' object has no attribute '_keepalive_e" <- this bug fixed in upstream https:/ /bugs.launchpad .net/ironic/ +bug/1496017