On 11/12/20 2:45 AM, Ben Pfaff wrote:
> Many of these could be replaced by "ovn-nbctl sync".  Some weren't
> really needed at all because they were adjacent to something that itself
> called sync or otherwise used --wait.  Some were more appropriately
> done with explicit waits for what was really needed.
> 
> I left some "sleep"s.  Some were because they were "negative" sleeps:
> they were giving time for something to happen that should *not* happen
> (in other words, if you wait for it to happen, you'll wait forever,
> unless there's a bug).  Some were because I didn't know how to properly
> wait for what they were waiting for, or because I didn't understand
> the circumstances deeply enough.
> 
> Signed-off-by: Ben Pfaff <[email protected]>
> ---

There are still tests that occasionally fail when I run the suite with
"-j20".

I'm pretty sure they're all related to the "--wait=hv sync" bug [0].

If I patch OVS to include the following IDL changes:
-
https://patchwork.ozlabs.org/project/openvswitch/patch/[email protected]/
-
https://patchwork.ozlabs.org/project/openvswitch/patch/[email protected]/

And if I patch OVN with the following commit which depends on the OVS
patches:
https://github.com/dceara/ovn/commit/c4926fbb386a3fa83c99618b6b12bcb4466fd1d6

Then the tests pass (or at least I'm not hitting this specific race
condition anymore).

Would it be possible to have a look at the OVS IDL patches above?  Once
those are merged I can cleanup and post the OVN patch.

Thanks,
Dumitru

[0] https://mail.openvswitch.org/pipermail/ovs-dev/2020-November/377129.html

_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to