[Kolab-devel] Kontact reliability - revisited

[ITSEF Admin]

Hi all,

I'm sorry to bring this up again, but given the increasing distrust of our 
users against Kolab/Kontact, I just *have* to come up with a solution...

We have some 35 users, all using Kontact proko2 (by now 2.1.7) on SuSE 10.0. 
Our home directories are NFS mounted, *however*, due to performance problems 
(issue1464), I switched to a combined setup early on. Combined means that 
$HOME/.kde/share/config and $HOME/.kde/apps/kmail/dimap are *not* on the NFS 
share, but on the local drive and then linked in using symbolic links[0]. I 
know that some Kolab developers frown upon NFS in this context (and for 
plausible reasons, too). However, given that the important directories are 
local, I think that NFS should not be the culprit here? Or am I missing 
something?

Anyway, there are increasing rumours (I know, I know...) about kontact being 
unreliable and so far, six users filed one or more reports about vanishing 
events. I followed up on a few of them and was able to find the following 
facts:

- Some cases were due to issue1498 (appointments vanish from view, but come 
back after a restart)
- In several cases, events were created but they somehow vanished from the 
server - they did not show up in the nightly backup of the same day. For some 
reason, they disappeared somewhere between creation and the backup time.
- The problems do not seem to be related to whether or not other people were 
invited - I've seen cases where the event had disappeared from the 
originator's calendar, but was still present in the attendee's calendar(s). 
In one case, I could even still find the confirmation answer of the attendee 
in the originator's trash, but the event itself was nowhere to be found in 
the originator's calendar or backups thereof.
- I had two cases where the agenda was emptied completely. As far as I 
understand the whole system, this should only happen if for some reason the 
local DIMAP cache is removed, right? In that case, I *might* tend to put it 
down to [0].

For the three most affected users, I've created an additional 24h hourly 
backup, however, that does not help if the user only realises that an event 
has disappeared days after the creation.

I myself am using the exact same setup, but have yet to loose a single 
appointment, despite intense use - one of the reasons why this is so 
difficult to handle... I still do not exclude user mistakes, to be honest, 
however, right now I can't quite see how a user can accidentally, without 
noticing, remove an appointment himself.

Also, I'd love to get rid of the scripting hacks. However, from what I gather 
from previous responses, if we continue to use NFS, all bets are off as far 
as the Kolab team is concerned, so we'd definitely have to consider moving to 
AFS - is that the official stance now?

As always, thanks in advance for all and any insights you might have for me. 
I'm really scratching my head here, especially, as I see no other reports 
about similar problems e.g. on kolab-users - so there is a chance we're doing 
something stupid. If only I knew... :-}

Regards,

Thomas


[0] this uses some ugly scripting hacks to enable people to log in on another 
machine, but that's a different kettle of fish. Hardly any of our users do 
log in on another PC, so, fortunately, this doesn't come into play a lot.