On Tue, Aug 07, 2012 at 02:38:48PM +0200, Jan Zelený wrote: > Dne úterý 07 srpna 2012 08:12:44, Stephen Gallagher napsal(a): > > On Tue, 2012-08-07 at 10:23 +0200, Jakub Hrozek wrote: > > > On Mon, Aug 06, 2012 at 12:48:14PM -0400, Stephen Gallagher wrote: > > > > On Mon, 2012-08-06 at 18:11 +0200, Jakub Hrozek wrote: > > > > > https://fedorahosted.org/sssd/ticket/1459 > > > > > > > > > > I noticed this while trying to reproduce #1364. The SRV servers were > > > > > just silently skipped because they were not marked as primary. > > > > > > > > Ack > > > > > > Dmitri raised an interesting point in the ticket. Our documentation seems > > > to suggest the _srv_ servers can be present in the backup list. > > > > > > I actually think we should just disallow that and only support host names > > > in the backup list (and I seem to remember that we said that the backup > > > server implementation would keep it simple and won't deal with SRV > > > servers at all). > > > > > > Supporting _srv_ records would mean all kinds of interesting corner cases > > > like handling duplicates between the primary and secondary list on the > > > fly, > > > because you don't really know what the SRV query would yield. > > > > > > We can't fix the documentation now that we're past the string freeze, > > > though.. > > > > We've got a couple other potential manpage issues that have come up as > > well. See https://fedorahosted.org/sssd/ticket/1461 for an example. > > > > I'm thinking it might be worth our while to make an announcement on > > Transifex that we're going to break string freeze this week because of > > some late-discovered issues and that we'll re-push the source strings on > > Friday. > >
Done. The Transifex announcement should land on the list shortly. > > With that in mind, I agree that we should probably restrict our usage to > > the primary servers at this time. Please recheck the code and make sure > > that's actually implemented that way as well. I'm CCing Jan directly to > > get his input. > > I already discussed this with Jakub on IRC. I agree that disabling _srv_ > support for backup server list is the most convenient solution. It's also the > best one if we want this to be fixed quickly. OK, I'll push the one-liner attached at the beginning of the thread to get the SRV support working in master again. Right now, it's not functional at all. > > However I don't think there will be a lot of corner cases we will have to > specifically handle in case we decide to do _srv_ support in backup server > list. For example the issue with one server being able to be both in primary > and backup list is there even if you don't use srv lookup. Fixing it should > not be that difficult and extending the fix to cover _srv_ as well will take > ~15 > copy-pasted lines from a brief overview of the code. > I'll be testing how the duplicate detection works and if whether _srv_ tag can be used in the backup list. I'm going to take a stab at how difficult it would be to put _srv_ into the backup list but my gut feelign is that we shouldn't bother if it's anything bigger than a trivial fix. It's very unlikely to be useful anyway. _______________________________________________ sssd-devel mailing list sssd-devel@lists.fedorahosted.org https://lists.fedorahosted.org/mailman/listinfo/sssd-devel
