Deployment silently failed when openvswitch can't start due to memory lack
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Fuel for OpenStack |
Fix Released
|
High
|
MOS Linux | ||
Mitaka |
Fix Released
|
High
|
MOS Linux | ||
Newton |
Fix Released
|
High
|
MOS Linux |
Bug Description
Steps:
1. Create env with Neutron Vlan network
2. Add two slaves: 1 controller and 1 compute
3. Move private network to enp0s5
4. Enable DPDK on enp0s5 for compute node
5. Configure HugePages for compute node: 2m = 5, 1g = 0, dpdk = 0
6. Run deployment
Actual result:
Deployment failed.
Puppet and astute log contain no errors. Only one warning in astute log about puppet-agent timeout on compute node.
Issue with openvswitch which can't start due to lack of memory:
EAL: Not enough memory available on socket 0! Requested: 64MB, available: 20MB
PANIC in rte_eal_init():
Cannot init memory
7: [ovs-vswitchd() [0x408aa3]]
6: [/lib/x86_
5: [ovs-vswitchd() [0x407289]]
4: [ovs-vswitchd() [0x505292]]
3: [/usr/lib/
2: [/usr/lib/
1: [/usr/lib/
Aborted (core dumped)
* Starting ovs-vswitchd
* Enabling remote OVSDB managers
root@node-2:~# echo $?
0
puppet freeze on starting openvswitch and don't report any error.
Expected result:
Puppet logs contain error which points engineer to issue with openvswitch and/or memory misconfiguration.
Env:
9.0 customm http://
description: | updated |
tags: | added: feature-dpdk |
Changed in fuel: | |
milestone: | none → 9.0 |
assignee: | nobody → MOS Packaging Team (mos-packaging) |
Changed in fuel: | |
status: | New → Confirmed |
Changed in fuel: | |
importance: | Medium → Low |
Changed in fuel: | |
status: | In Progress → Fix Committed |
tags: | added: verification-needed |
Changed in fuel: | |
assignee: | MOS Linux (mos-linux) → Maksym Strukov (unbelll) |
Changed in fuel: | |
assignee: | Maksym Strukov (unbelll) → nobody |
Changed in fuel: | |
assignee: | nobody → MOS Linux (mos-linux) |
Patch