On Thu, Jul 23, 2020 at 10:23 AM Florian Schmid <[email protected]> wrote: > > Hello Yedidyah, > > thank you for this great answer. > > I will answer in the text below. > > BR Florian > > > ----- Ursprüngliche Mail ----- > Von: "Yedidyah Bar David" <[email protected]> > An: "Florian Schmid" <[email protected]> > CC: "Tomas Golembiovsky" <[email protected]>, "Sandro Bonazzola" > <[email protected]>, "users" <[email protected]> > Gesendet: Donnerstag, 23. Juli 2020 08:37:21 > Betreff: Re: [ovirt-users] Re: qemu-guest-agent on Ubuntu doesn't report FQDN > > On Wed, Jul 22, 2020 at 5:34 PM Florian Schmid via Users > <[email protected]> wrote: > >> > >> Hi, > >> > >> after digging a bit deeper, it looks like it is the problem with the > >> qemu-guest-agent. > >> > >> It does only report the hostname and nothing more. It uses this function: > >> g_get_host_name () > >> > >> This function always returns the value in /etc/hostname and this is > >> normally the short name of the VM without the domain part. > >> > >> It looks like, that the ovirt-guest-agent made this different, > > > >Indeed, and from checking the git log, it seems like it did this since > >the very first commit - already then, > >ovirt-guest-agent/GuestAgentLinux2.py had: > > def getMachineName(self): > > return socket.getfqdn() > > Correct, this is what I wanted back. > > > > >> but this is not working anymore with python 3. > > > >If in "this" you refer to ovirt-guest-agent, then it's deprecated: > > > >https://bugzilla.redhat.com/show_bug.cgi?id=1672732 > > Yes, I know. Now using the QGA with oVirt 4.3 reports only the short hostname. > > > > >> > >> There was a recent patch for qga -> > >> https://bugzilla.redhat.com/show_bug.cgi?id=1845127 > > > >This bug seems to discuss something else, not directly related to your > >own issue. > > > >> but this won't help me, because even when this patch would add the FQDN to > >> oVirt back, there won't be a package for this for Ubuntu 20.04 and > >> probably also not for RedHat/CentOS 8. > > > >Not sure what you mean here. The bug is on qga, and fixing it (or your > >own issue) is unrelated to oga's deprecation. > > I wanted to say, that this change might also impact the reported hostname, > but I don't think so... > > > > >Your issue seems to be, to me: > > > >1. oga used to report the FQDN, as returned by python's socket.getfqdn() > >2. qga returns something else (and this something else might be > >changed, following above bug, but likely not to what you want). > >3. oVirt now uses qga instead of oga, thus changing its past behavior. > >4. You want the old behavior back - basically, claiming this is a regression. > > Yes, exactly. > > > > >If so, then: > > > >1. You are welcome to open a bug about this, on qga. > >2. Your request *might* be rejected, on the ground of breaking > >compatibility for existing/old users of qga (say, using virt-manager > >or whatever other virt tool, without oga installed) > > I'm 100 % sure, that they will reject this. > > > > >Alternatively, or if this bug is rejected, you can open two new bugs: > > > >1. one on qga, to provide the fqdn (using, hopefully, logic similar to > >python's getfqdn, although qga is written in C) > > Possible, but this won't help me a lot, because even if they add a new > function to qga, oVirt would need to be changed too, to access this function > instead of the one it is using now. > > >2. other on the oVirt engine, to use this new functionality of qga > >instead of the existing one. > > Yes. > > > > >You also have another alternative - just adapt your machines to have > >the fqdn as the hostname. I personally think this is the best way to > >go. Have 'hostname' return the FQDN you want, and only use 'hostname > >-s' where you really want it to be short. How do you set the hostnames > >of your machines? > > This is what I don't know, if this has some drawbacks. > I have checked this on internet, but haven't find a lot about it, what is > digging deeper. > > Maybe someone here has some experience with using FQDN for hostname?
I use this on all my machines (CentOS/RHEL ones, anyway) and all seems ok. I do recommend, obviously, that you do your own research/testing. > > I can live with such a solution, when it doesn't have big drawbacks... The only actual drawback I can think of is that the hostname is limited to 64 chars, whereas the FQDN can be up to 255 chars. So if you want a longer FQDN, you can't use it as the hostname. Another obvious drawback is that applications that use the full hostname for reporting, as opposed to explicitly using 'hostname -s', will now have the FQDN in their reports, which you might find too long etc. I personally consider this an advantage, not a drawback. Memory and disk space are cheap these days, but time is still expensive - if you get a report about "servera" instead of "servera.somedomain" and it causes you to spend time understanding what server this is, it's more waste, IMO, than having "servera.somedomain" everywhere. But that's obviously a personal matter... Best regards, -- Didi _______________________________________________ Users mailing list -- [email protected] To unsubscribe send an email to [email protected] Privacy Statement: https://www.ovirt.org/privacy-policy.html oVirt Code of Conduct: https://www.ovirt.org/community/about/community-guidelines/ List Archives: https://lists.ovirt.org/archives/list/[email protected]/message/OKEXNT7SDNP4MVH2JG23QHTZN4AN3JME/
