[NETVIRT-272] ARP thread is sleeping 2s each time MAC migration occurs Created: 14/Nov/16  Updated: 04/Dec/16  Resolved: 04/Dec/16

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

Type: Bug
Reporter: Koby Aizer Assignee: Olga Schukin
Resolution: Done Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified
Environment:

Operating System: All
Platform: All


External issue ID: 7170

 Description   

If there an entry of some IP with MAC resolution X, and this entry is updated to MAC resolution Y, then ArpNotificationHandler sleeps for 2s.

From mailing list:

  • ArpNotificationHandler is sleeping for 2s each time it gets an ARP request/reply which changes a resolved MAC address in the cache.
    Even though this scenario sounds rare, current ext-gw resolve implementation creates packets that cause that all the time, and therefore we hit this code block quite a lot.

We see two sleeps introduced in the entire ArpNotificationHandler and they are applicable only when there is a MAC migration seen for an existing IP address.. By MAC migration, means the ODL controller is receiving that same IP in packets now from a different interface.

The sleeps are introduced because when a MAC migration happens for an existing IP, there is a VRF Entry refresh lifecycle that gets triggered for that IP.

In order to make sure VRF removal is completely processed for old interface, before we do a VRF add for the same IP Address for the new interface a sleep was added between the removal and add operations..


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