Fuel 6.1 + mellanox plugin - Kernel panic on controllers
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Fuel for OpenStack |
Won't Fix
|
High
|
Fuel Partner Core |
Bug Description
Under a common and simple situation, one tenant is able to crash the controllers (kernel panic) when generating network traffic between two instances having Floating IPs.
Steps to reproduce:
1) Create two instances (I've used TestVM) on the same subnet with a security group allowing SSH traffic;
2) Check that one instance can connect the other via SSH (and vice-versa) over the DHCP IP addresses obtained in that subnet;
3) Associate a Floating IP to one instance;
4) Connect via SSH from the instance having private IP to the other having the Floating IP: it works as expected;
5) Associate another Floating IP to the second instance;
6) The SSH connection from one instance to the other, using its Floating IP, causes a kernel panic on one controller;
7) Iterating the same steps crashes another controller. At this point the system become unusable with most services unreachable (horizon, instances with floating IPs connectivity go offline...).
Workaround:
Nothing found to prevent the wrong behaviour.
A reboot of the controllers is needed to get control back.
Description of the environment:
- Fuel 6.1 community + Mellanox plugin 1.0
- CentOS 6.5 Juno with kernel 2.6.32-
- Ceph storage
Network model:
- Neutron + VLAN segmentation networking over InfiniBand connectivity
Changed in fuel: | |
assignee: | nobody → Fuel Partner Core (fuel-partner-core) |
Changed in fuel: | |
milestone: | none → 6.1-updates |
importance: | Undecided → High |
tags: | added: area-plugins |
Changed in fuel: | |
status: | New → Confirmed |
We no longer support MOS5.1, MOS6.0, MOS6.1 Security fixes to MOS9.2.
We deliver only Critical/Security fixes to MOS7.0, MOS8.0.
We deliver only High/Critical/