[Solved in 11.4] 11.3 Outlook connector can't connect

[tomster]

Word!

It's not that this thread is the only one dealing with problems on V11.3 and so far not much feedback from the Scalix guys. Yeah, a hosts-entry, great...

We've been desperately waiting for 11.3 and it's loudly announced Outlook 2007 compatibility. Now we try to run 11.3 and nothing but hassle with my new users all having been "upgraded" to Outlook 2007. That might not be Scalix' fault in the first place, but shouldn''t you have had any BETA testers bringing that issue up within the first 2-3 guys installing the new connector?

OK, OK I'm on the community rail so far, but as of now I have to agree with my company's budgeteers' opinion on not spending a dime (let alone some grand) on a SBE licence at this point in time. And I'm really a strong defender of your piece of Software here!

Ahh, and the dropped SyncML-Feature would hit the same line here. Active Sync, NotifyLink, etc... Why not stick with Exchange then? The cost-wise advantage of Scalix over Exchange is gradually diminishing, don't you think?

I really appreciate Florian's efforts on resolving any problem brought up here in the forums, especially since we're mostly all community users here (paying back only in feed-back rather that mean green). But don't you think it's about time to get Scalix to a "working state" covering at least the most common setups (Scalix + Outlook 200X)?
Just my 2 cents (not wanting to hi-jack this thread)...

TOM

[Valerion]

I must admit most of my migrations went very smoothly, actually. And 11.3.0 seems very stable for me (WinXP, OL2k3). The only issue I found so far was the DNS issue les reported in a different thread. Hosts file entries are not needed for me at all.

On my corporate email I set up my scalix-default-mail DNS entry to point to mail.company.com, which points to the correct IP address, and my Outlook connector picked that up fine when I created my profile. This was with both SSL enabled and disabled. I later closed the non-SSL port on the firewall, so everyone is forced to use SSL.

As to wireless syncing, Florian has been making comments regarding a possible solution there, but he has also stated it will not be ready for 11.3.0.

[les]

I must admit most of my migrations went very smoothly, actually. And 11.3.0 seems very stable for me (WinXP, OL2k3). The only issue I found so far was the DNS issue les reported in a different thread. Hosts file entries are not needed for me at all.

On my corporate email I set up my scalix-default-mail DNS entry to point to mail.company.com, which points to the correct IP address, and my Outlook connector picked that up fine when I created my profile. This was with both SSL enabled and disabled. I later closed the non-SSL port on the firewall, so everyone is forced to use SSL.

As to wireless syncing, Florian has been making comments regarding a possible solution there, but he has also stated it will not be ready for 11.3.0.

For me also, everything else went smoothly and the general increase in speed across SWA, Outlook and IMAP is impressive and worth the upgrade. I just wish it didn't have the DNS lookup issue, that would have saved me a return visit the next day.

For the DNS Issue i have file a bug report.....

https://bugzilla.scalix.com/show_bug.cgi?id=16517

[Valerion]

I think what saved me effort there is that I try as fart as possible to stay away from using different domains inside and outside my corporate networks. As long as the domain of the server and the pointer for scalix-default-mail is in the same domain, I didn't have issues.

But I agree, this does sound like a bug to me. Thinking about it again, I did have issues in the case where the DNS is structured as follows (domains left out):

scalix-default-mail CNAME -> mail CNAME -> real-host-name

However, I changed it to

scalix-default-mail CNAME -> mail A -> machine IP
The hostname stayed as real-host-name.domain.com.

In the first case the name appearing in the profile creation screen was real-host-name. In the second case it was mail, as expected. There is definitely an issue with how far the resolution gets done.

I wonder if the client doesn't make a connection to the server, which then returns its own host name, and have that populated into the profile somehow? Of course it should return the value from the mailnode mappings table, but it may attempt to completely resolve that to a IP and see what the last lookup is it did?

Of course that's all idle speculation until the Scalix developers gets back to us.

[les]

I think what saved me effort there is that I try as fart as possible to stay away from using different domains inside and outside my corporate networks. As long as the domain of the server and the pointer for scalix-default-mail is in the same domain, I didn't have issues.

But I agree, this does sound like a bug to me. Thinking about it again, I did have issues in the case where the DNS is structured as follows (domains left out):

scalix-default-mail CNAME -> mail CNAME -> real-host-name

However, I changed it to

scalix-default-mail CNAME -> mail A -> machine IP
The hostname stayed as real-host-name.domain.com.

In the first case the name appearing in the profile creation screen was real-host-name. In the second case it was mail, as expected. There is definitely an issue with how far the resolution gets done.

I wonder if the client doesn't make a connection to the server, which then returns its own host name, and have that populated into the profile somehow? Of course it should return the value from the mailnode mappings table, but it may attempt to completely resolve that to a IP and see what the last lookup is it did?

Of course that's all idle speculation until the Scalix developers gets back to us.

I agree, but 11.2 and before never worked like that. And i was always led to believe that upon setup the connector did a dns lookup based on whatever the pc used for dns.

In the past on sites where i hadn't had a scalix-default-mail CNAME i got a blank field where i could type in any name. On sites where i did have a CNAME for scalix-default-mail, i got a proper hostname returned from dns, whether it be a "mail.lan" or "mail.mydomain.com" based on what the pc could lookup and it worked.

It certainly didn't work the other day in 11.3 even though the pc could lookup the servername without problem. And when i typed in "mail.lan" it would not go past the servername prompt, continually saying the servername was not found. That's a bug in my book ;)

[jaime.pinto]

Our DNS is *perfect*, no CNAME, no internal/external mix, no dual names, no port redirect, no local hosts file on PC, etc. No firewall or port filtering on the client. Forward and reverse lookup on the client return exactly what it should. So, what is going on inside 11.3 connector?

[Valerion]

You should have at least one CNAME, scalix-default-mail, pointing to the server name.

What are you seeing, and what does the profile creation dialog show for a server name?

[jaime.pinto]

Well, this is not it.
Our DNS servers have been revised to publish 'scalix-default-mail' as a CNAME for the FQDN (server.domain.com), which until now has been the *only* one used on the scalix server as well as throughout all client systems, external/internal and DMZ.

When I try to create a new profile the "Scalix Login Information" window presents me a form with 3 *empty* fields
Server Name:
User name:
Password:

I'm absolutely sure about User/Password fields.
For server name I entered the FQDN (server.domain.com) as well as the CNAME (scalix-default-mail.domain.com)
An error windows comes up Server not found on network
We have no problems accessing SWA from the same machine with either FQDN or CNAME

Hi Florian
Any chance you could take personal ownership of this issue/.bug, since it affects a very large number of end users OL2003/2007 once the 11.3.0 connector upgrade is applied.? I'd be happy with an acknowledge/admission that the current version seems to be broken and scalix dev-team is checking things out further under controlled conditions in your own lab, and will release a new version asap. I'll just hold up the upgrade.
Thanks

PS: in the morning I'm removing the ridiculous CNAME as well.

[dyna]

CNAME trick did nothing for me either.

Well atleast good to see others agree it's a DNS issue and that the name set in the connector is most likely not retrieved via DNS lookup.

[les]

CNAME trick did nothing for me either.

Well atleast good to see others agree it's a DNS issue and that the name set in the connector is most likely not retrieved via DNS lookup.

If you look back through the posts you'll see a bug link i filed. In that Florian explains what is happening and what has changed.

Quoting Florian.....

the behaviour in 11.3 has changed for SSL support. Due to this fact, the
hostname needs to be fully consistent in what's used in hostname --fqdn, Scalix
Connect profiles, hostname resolution on the outlook side and omshowmnmp. No
exceptions allowed. It is hard to devise where exactly this fails in the
situation below and if help is required, please do involve Scalix tech support.
The behaviour of Scalix Connect is as-designed, what happens is that on
successful Outlook connection, the servername in the profile is replaced with
the servers own notion of the servername, i.e. omcheckgc -h output (which, on a
single instance system will be equal to hostname -f).

We will come up with a more complete solution for outlook hostname/ip address
connection handling, however, this is a bigger task and will not happen before
the next full release.

Again, I believe the behaviour described is not a bug, but just a result based
on inconsistent host naming and specific local configuration.

[jaime.pinto]

On the server:
hostname --fqdn EQUAL omcheckgc -h EQUAL hostname -f EQUAL server.domain.com

This is the same result you get from forward and reverse DNS lookups on the DMZ, internal or external to the firewall.

But omshowmnmp:
server server.domain.com
internet

I'm sure the "server" part comes from the /etc/hosts file, and should not matter. But where is this "internet" line coming from? Does it really matter, ie, would this be the issue keeping the connector 11.3. from working on the client side?

[Valerion]

server is the mailnode on the machine that users are created in. Use omshowmn to verify this. All mailnodes should have entries in omshowmnmp. internet (and internet,tnef) is a pseudo-mailnode used for outgoing emails to the internet and incoming emails.

[SidebandSamurai]

As a test how about trying to replace the FDQN with the ip address.

Instead of typing scalixserver.companyname.com use 123.456.789.012 or what ever the IP address is of the server, just to see if the connector will connect successfully.

if the connector connects, and all other names resolve correctly, (both forward and reverse) then we might have a connector issue. Also starting with a blank host file on the local pc would help. I also know though not required if you make a change to the host file, sometimes its a good idea to reboot the work station which will flush the dns cache of the workstaton.

Sideband Samurai

[tomster]

As much as I remeber you simply CAN'T fill in an ip address. You need to put a FQDM there.

I run Scalix as an internal mail server. This means that our "real" mail server is hosted by our provider. Scalix fetches all email from that server and puts the email in each local user's mailbox. I simply was too lazy so far to read into setting up Scalix as our real mail server, dropping the provider's one.

But I assume this is a common set-up being used by numerous other Scalix disciples.

Scalix is run on a server that is equipped with 2 NICs. One LAN device and one WAN device. I would assume that all internal clients would prefer to access Scalix via the LAN interface. So I add an entry in /etc/hosts on the Scalix machine like:

Code: Select all

192.168.0.123 mail.server.com

Assuming that the Scalix machine is not the default gateway (in my case it's 192.168.0.1) for the LAN accessing the internet, each client needs a separate entry in its /etc/hosts also resolving mail.server.com to 192.168.0.123. Otherwise the gateway would try to find mail.server.com outside the LAN. This is in may case necessary because I use a out-of-the-box router on which I can't edit any /etc/hosts file on.
To make Scalix also accessible from the WAN side you would need to have mail.server.com being (publically) resolved to Scalix' WAN ip address. That's the easier part since this is done by numerous external nameservers.

I sort of understand Florian's explanation (at least I think I understood) in some way, but wouldn't that imply that Scalix can necessarily only being accessed by its FQDN? And from my understanding resolving that name should always point to the WAN IP. The entries in the client's /etc/hosts only provide a workaround to not route internal requests via the default gateway. But they then have to be set on each internal client, right?

I do want to dig into setting up a local DNS service here, because in my case (I use 2 different internet connections) this can build up to an administrative nightmare since I want all servers to be accessible from both connections.

Sorry for writing that much without coming to my actual point, but I think you need to know my intentions to be able to answer my question:

My /etc/hosts file on the Scalix machine states:
192.168.0.123 mail.server.com server.local

I did hostname -f and it shows:
mail.server.com

Strangely a omshowmnmp shows:
server server.local

Did I mess up my hosts file? I though that multiple aliases shouldn't be a problem, right?

I am not exactly a network geek so I'd appreciate a little help here.

--edit--
OK, the hostname thingy seems solved. I think I messed up something when changing FQDN. Redid it following the wiki, now there's the correct showommnmp output.

Anyway, a little help on the DNS setup is happily welcome.
Greets
TOM

[mikevl]

Hi

Your /etc/hosts should show
192.168.0.123 mail.server.com mail

Please do not get confused with default gateway and DNS.
In your workstation ip configuration regardless if it is using DHCP or manual DNS entry, it will point to a DNS server. Hopefully the DNS server your workstation points to will be on your LAN.

Your DNS server should hold
An "A" record for the domain server.com (I make mine server.local cause this does not interfear with external lookups),

mail 192.168.0.123 (under your domain the entry will be ammended to mail.server.com)
Then you need a CNAME record scalix-default-mail which points to your server i.e.
scalix-default-mail > mail.server.com

Your default gateway defines how trafic which does not have a LAN address finds its way out of the LAN, maybe to the internet or the next router. For LAN traffice between Workstations and servers a default route is not neccessary.

Hope this helps

Mike