[NETVIRT-268] Floating IP rules installed with wrong actions Created: 13/Nov/16  Updated: 19/Oct/17  Resolved: 12/Jan/17

Status: Resolved
Project: netvirt
Component/s: General
Affects Version/s: Boron
Fix Version/s: None

Type: Bug
Reporter: ofer barber Assignee: Unassigned
Resolution: Done Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified
Environment:

Operating System: Linux
Platform: All

Attachments: Zip Archive bug_netvirt.zip    
External issue ID: 7165

 Description   

I have a setup with 3 OpenStack controllers, 1 Karaf and 2 compute nodes.
I have constructed the following topology:
two private networks(each configured with vxlan) and one external network connected to one router.
each private network has one VM.

Ping between vms work fine.
Ping to floating IP (of either vms) doesn't work.

I have investigated the first floating ip, 10.83.5.112, which is used by a vm on compute 1. In the ovs dump flows of compute 1, I see the following flow:

cookie=0x8000003, duration=6378.900s, table=21, n_packets=66, n_bytes=5868, priority=42,ip,metadata=0x222e0/0xfffffffe,nw_dst=10.83.5.112 actions=write_actions(set_field:0x11245->tun_id,output:5)

port number 5 is a tunnel port and not a "vm" port.

Dump model of config/odl-fib:fibEntries/ shows:
{
"destPrefix": "10.83.5.112/32",
"label": 70213,
"nextHopAddressList": [
"10.83.6.5",
"10.83.6.6"
],
"origin": "s"
},

10.83.6.5 is compute1 IP, 10.83.6.6 is compute2 IP.

 Comments   
Comment by ofer barber [ 13/Nov/16 ]

Attachment bug_netvirt.zip has been added with description: Karaf, OVSs and model

Comment by ofer barber [ 13/Nov/16 ]

Karaf version:
distribution-karaf-0.5.2-20161109.002212-2

Comment by Tomer Pearl [ 15/Nov/16 ]

It seems like that this issue happens when reusing the floating ip between vms, without deleting it from the odl-fib:fibEntries model, in this case a new entry is appended to nextHopAddressList.

When i'm trying to reproduce this issue from horizon, after dissociating the floating ip from a vm, the relevant entry in deleted from odl-fib:fibEntries model, and when i'm associating the floating ip again, a relevant entry is created correctly in the model.

Comment by dotans [ 23/Nov/16 ]

I have a Newton setup.
I created NET,router and VM one time.
Then, in a repetitive test: create floating IP, associate, ping to FIP (ping should pass),delete FIP and ping to FIP (ping should fail).
The same happened to me 3 times during night (cycles 11,512,1227).
logs are at:
\\10.0.1.216\pub\DotanS\Bugzilla_7165

Comment by dotans [ 02/Jan/17 ]

The scenario was reproduced again with a simple creation of network,vm,router and association of FIP to vm.
I deleted the floating ip, router,vm and network but didn't delete the external network.
When creating again (private network,router, vm and association of FIP)- this bug reproduced.
I added another VM and its FIP was pingable (no reproduction).
After deleting all configuration:private net,external net,router,FIP and VM - the bug wasn't reproduced again

Comment by Koby Aizer [ 12/Jan/17 ]

Analyzing this further with Dotan, we got to a conclusion that this is fixed by: https://git.opendaylight.org/gerrit/#/c/50110/

Generated at Wed Feb 07 20:21:08 UTC 2024 using Jira 8.20.10#820010-sha1:ace47f9899e9ee25d7157d59aa17ab06aee30d3d.