DNS resolving is very slow with the procedure of "Deploying Internal DNS with OVN" I tested "Deploying Internal DNS with OVN" with following the procedure. It looks work but it's very slow to resolve the instance name. It takes 20~60 sec. However, when I added the gateway as DNS server to the subnet, it has been improved to less than 1 sec. And, I'm wondering if this is the proper setting for this feature. Anyway, I think the document is not sufficient to use this internal DNS feature because it's too slow to use it. Reported by: rhn-support-migawa https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/16.1/html/networking_with_open_virtual_network/deploying-dvr-ovn#annotations:dd8d2208-21a4-4a47-8095-664b6129c813
Hi, Based on Eduardo's conclusion at the end of Comment 4, it sounds like there is no action required for the RHOSP customer docs on this issue. Therefore, I am closing this BZ. If I have missed something, please comment and reopen this BZ. Thanks, --Greg
Hi Greg, Thank you for taking care of this. I think it's OK to close this bug because we already opened the new bug BZ#1946662 to handle this issue. And, I will create a KCS for it. Best Regards, -- Masayuki
@Eduardo, @Greg, the OVN bug was closed, BUT: the new option for DNS record to "own" the domain should be set by CMS (or user). If we think that this should be the OVN ml2 driver behavior, then there's a missing piece - neutron ml2 driver integration to set the new option. Should we report a new Jira to track this work?
Hi @Ihar, Thanks for raising this issue. I would vote for working this into the ML2/OVN default behavior. One less knob for customers to turn for something that should be already the default. Best, --Greg R.
Hi Ihar and Greg, After taking a look at the description of the core ovn's upstream patch [1], it seems to me the option `ovn-owned` should be set to True by default. I am thinking about it and I can't see any case where setting it to True could be problematic. On the other hand, this would be a change of behavior, hence, rather a (small) feature than a bug, right? We would need a few tests covering the new behavior (e.g. networks with both IPv4 and IPv6 subnets and networks with only one of them). It would only go to RHOSO18? Or RHOSP17.1 too? BR, Eduardo [1] https://patchwork.ozlabs.org/project/ovn/patch/20231001192704.3107450-1-mheib@redhat.com/
I think we can start with a RHOSO 18 Jira and consider the complexity of the implementation for 17.x later. Would you mind reporting a Jira for this work to happen in neutron? Thanks.
(In reply to Ihar Hrachyshka from comment #16) > I think we can start with a RHOSO 18 Jira and consider the complexity of the > implementation for 17.x later. Would you mind reporting a Jira for this work > to happen in neutron? Thanks. I have opened this ticket: https://issues.redhat.com/browse/OSPRH-10758