[L2SWITCH-31] Loop Remover only works 1 time after controller starts Created: 23/Jan/15 Updated: 19/Oct/17 Resolved: 22/Jun/15 |
|
| Status: | Resolved |
| Project: | l2switch |
| Component/s: | General |
| Affects Version/s: | unspecified |
| Fix Version/s: | None |
| Type: | Bug | ||
| Reporter: | Luis Gomez | Assignee: | Unassigned |
| Resolution: | Done | Votes: | 0 |
| Labels: | None | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Environment: |
Operating System: Linux |
||
| External issue ID: | 2629 |
| Description |
|
Loop remover only works first time you bring mininet after controller starts, second and sucesive times does not work. To reproduce: 1) Start controller 2) Start mininet in ring or regular topology like linear 3) Check port inventory and see loopremover working: <node-connector> 4) Stop mininet 5) Start mininet again (same topology used in 2) 6) Check port inventory, now there is no loopremover in any port 7) In case of ring topology this will cause packet loop BR/Luis |
| Comments |
| Comment by Luis Gomez [ 27/Jan/15 ] |
|
I will add the comment that in the case of a ring topology, when the loop remover is not present and a host sends a ping, this produces a packet storm not just in the data plane but also on the control plane (l2switch flood flows send traffic to controller). This produces all kind of issues in controller like high CPU, dummy switches after stopping mininet, no host discovery, etc... Only workaround I have seen is to restart controller. BR/Luis |
| Comment by Evan Zeller [ 14/Apr/15 ] |
|
See: https://git.opendaylight.org/gerrit/#/c/18218/ In addition to the loss of StpStatus on NodeConnectors it seems this bug has scale impacts as well. Before: After: sudo mn --topo tree,8 (255 nodes) at 511 nodes (--topo tree,9) the exception "DOMRpcRouter-listener-0" java.lang.IllegalArgumentException: Route Identifiers must be supplied" reappears and some pings begin to fail. |
| Comment by Luis Gomez [ 15/Apr/15 ] |
|
And you made these perf tests using master code? there are some issues with openflow topology mgr that impacts the l2switch as you can see in the CI test: |