[Samba] Van Belle Repo - 4.10 for Bionic

Sun Mar 31 13:30:23 UTC 2019

On Sun, 31 Mar 2019 14:57:32 +0200
Viktor Trojanovic <viktor at troja.ch> wrote:

> On Sun, 31 Mar 2019 at 14:45, Rowland Penny via samba
> <samba at lists.samba.org> wrote:
> >
> > On Sun, 31 Mar 2019 14:29:32 +0200
> > Viktor Trojanovic <viktor at troja.ch> wrote:
> >  
> > > On Sun, 31 Mar 2019 at 14:22, Rowland Penny via samba
> > > <samba at lists.samba.org> wrote:  
> > > >
> > > > On Sun, 31 Mar 2019 13:55:18 +0200
> > > > Viktor Trojanovic via samba <samba at lists.samba.org> wrote:
> > > >  
> > > > > On Sat, 30 Mar 2019 at 17:45, L.P.H. van Belle
> > > > > <belle at bazuin.nl> wrote:  
> > > > > >
> > > > > > you can upgrade from 4.7 to 4.9.
> > > > > > but you must run samba-tool dbcheck
> > > > > > that should work also. after the check run samba-tool
> > > > > > dbcheck --reindex  
> > > > >
> > > > > I thought that maybe indeed I did something wrong the last
> > > > > time so I followed your advice. Unfortunately, it broke my DC
> > > > > again. (Luckily, I made a snapshot before doing this so I
> > > > > should be able to easily revert).
> > > > >
> > > > > So I went through with the update from 4.7.4 to 4.9.4, and
> > > > > after the update I ran:
> > > > >
> > > > > # samba-tool dbcheck
> > > > > Checking 322 objects
> > > > > Checked 322 objects (0 errors)
> > > > >
> > > > > # samba-tool dbcheck --reindex
> > > > > Re-indexing
> > > > > completed re-index OK
> > > > >
> > > > > # ldbedit -H /var/lib/samba/private/sam.ldb
> > > > > (shows contents of sam.ldb, closing the file)
> > > > >    # 0 adds  0 modifies  0 deletes
> > > > >
> > > > > # reboot
> > > > >
> > > > > # journalctl -xe
> > > > >
> > > > > Mar 31 11:43:52 dc1 samba[224]: task[dnsupdate][224]:
> > > > > /usr/sbin/samba_dnsupdate: ERROR(runtime):
> > > > > uncaught exception - (9711,
> > > > > 'WERR_DNS_ERROR_RECORD_ALREADY_EXISTS') Mar 31 11:43:52 dc1
> > > > > samba[224]: task[dnsupdate][224]: [2019/03/31 11:43:52.021002,
> > > > > 0] ../lib/util/ util_runcmd.c:327(samba_runcmd_io_handler)
> > > > > Mar 31 11:43:52 dc1 samba[224]: task[dnsupdate][224]:
> > > > > /usr/sbin/samba_dnsupdate:   File "/usr/lib
> > > > > /python2.7/dist-packages/samba/netcmd/__init__.py", line 177,
> > > > > in _run Mar 31 11:43:52 dc1 samba[224]: task[dnsupdate][224]:
> > > > > [2019/03/31 11:43:52.021202,  0] ../lib/util/
> > > > > util_runcmd.c:327(samba_runcmd_io_handler)
> > > > > Mar 31 11:43:52 dc1 samba[224]: task[dnsupdate][224]:
> > > > > /usr/sbin/samba_dnsupdate:     return self.
> > > > > run(*args, **kwargs)
> > > > > Mar 31 11:43:52 dc1 samba[224]: task[dnsupdate][224]:
> > > > > [2019/03/31 11:43:52.021375,  0] ../lib/util/
> > > > > util_runcmd.c:327(samba_runcmd_io_handler)
> > > > > Mar 31 11:43:52 dc1 samba[224]: task[dnsupdate][224]:
> > > > > /usr/sbin/samba_dnsupdate:   File "/usr/lib
> > > > > /python2.7/dist-packages/samba/netcmd/dns.py", line 945, in
> > > > > run Mar 31 11:43:52 dc1 samba[224]: task[dnsupdate][224]:
> > > > > [2019/03/31 11:43:52.021539,  0] ../lib/util/
> > > > > util_runcmd.c:327(samba_runcmd_io_handler)
> > > > > Mar 31 11:43:52 dc1 samba[224]: task[dnsupdate][224]:
> > > > > /usr/sbin/samba_dnsupdate:     raise e
> > > > > Mar 31 11:43:52 dc1 samba[224]: task[dnsupdate][224]:
> > > > > [2019/03/31 11:43:52.047043,  0] ../source4/d
> > > > > sdb/dns/dns_update.c:330(dnsupdate_nameupdate_done)
> > > > > Mar 31 11:43:52 dc1 samba[224]: task[dnsupdate][224]:
> > > > > ../source4/dsdb/dns/dns_update.c:330: Faile
> > > > > d DNS update - with error code 26
> > > > >
> > > > > Does anyone have an idea what causes these errors?
> > > > >
> > > > > - Viktor
> > > > >  
> > > >
> > > > Yes, it appears that the change to the GUID index mode that
> > > > first appeared in 4.8.0 is the culprit. I don't think anybody
> > > > really knows why it fails, but it is believed to be that 4.9.x
> > > > thinks the GUID index mode is being used, but the database is
> > > > still the 4.7.x type.
> > > >
> > > > There have been a few workarounds suggested, but, to my
> > > > knowledge, the only real upgrade path is to 'walk' up the
> > > > versions. From 4.7.x upgrade to 4.8.x, run this so the database
> > > > is changed to the GUID index mode and then upgrade to 4.9.x or
> > > > higher. 
> > >
> > > That was the last information I had. I thought something changed
> > > after seeing Andrew comment in the bug report
> > > (https://bugzilla.samba.org/show_bug.cgi?id=13760) and Louis
> > > confirming that this should actually work.
> > >
> > > Do you happen to know how to do the walk on Ubuntu? If I
> > > understand this correctly, unless I want to compile myself (and I
> > > don't), I need packages, and neither Ubuntu nor Louis provide
> > > packages for 4.8. Which is why I'm thinking to upgrade by just
> > > joining a new DC based on Samba 4.9.4. Or wait for Louis to
> > > publish 4.10 as this should, in my understanding, make it
> > > possible again to upgrade properly.
> > >
> > > - Viktor  
> >
> > It looks like the fix is in 4.9.5 & 4.10.0 and Louis has issued
> > Samba 4.9.5 packages ;-)
> >  
> 
> Not sure I follow.. I only just today set up Louis' repo anew and
> 4.9.4 is what I got so I'm quite certain that's the newest package
> he's currently offering, for Ubuntu at least.

Sorry, you are correct, 4.9.4 is the latest available from Louis's
Ubuntu repo, but 4.9.5 is in the Debian one.

You can do one of three things:
Wait until Louis gets the time (and inclination) to update the Ubuntu
repo.
Set up Debian stretch in a VM and add this as a 4.9.5 DC to see if this
fixes the problem.
Do nothing ;-)

> 
> Also, I'm actually still not clear if an update to 4.10 will really
> solve the situation. If I read the bug report correctly, all 4.10 does
> is run dbcheck --reindex for you. If that doesn't work for me if I do
> it manually, why would it work if it was done automatically? I had the
> impression, however, that the upgrade I did on the other machine
> (which I had to shut down quickly thereafter) from 4.7.4. to 4.10 went
> without issues. 

At the moment, this appears to be the fix and unless it is tested and
found not to work, there will not be another fix. It could be that the
reindex is run before Samba is actually started (just guessing here)
and this makes the fix work. If you don't try it, you will never
know ;-)

>In a nutshell, I'm utterly confused. :-D

You and me both ;-)

Rowland