How not to install Exchange: Help please!

[Mr.Ben]

Good Morning,

 

First a story:

 

I've just started here and between the previous two Network Managers the Exchange 'Upgrade' to 2010 seems to have been bodged.

 

First Problem is the lack of RAM in the Mailbox Servers (16GB for 1400 Users) in each. An easy fix.

 

However it seems the DAG group has also been set up incorrectly, leading to database migrations that cause some interesting problems.

 

Basically the DAG group was set up to replicate on the External Network rather than it's own internal one, So I have corrected this and the Databases seem to re-seed.

 

As soon as they are 'Healthy' though they drop out with the following Error in Disconnected and Resynchronizing:

 

The Microsoft Exchange Replication service was unable to perform an incremental reseed of database copy 'Email Database\EMAIL' due to a network error. The database copy status will be set to Disconnected. Error An error occurred while communicating with server 'EMAIL2'. Error: An operation was attempted on something that is not a socket

 

Any ideas?

 

My Google-Fu brings up nothing...

 

Thanks in advance.

 

Ben

Edited September 21, 2012 by Mr.Ben

[sukh]

Need to more clear on the DAG setup and the errors...

How many DAG's, nodes, domains, Net cards in each DAG, teamed or not? Replication network on vlan? any ACL's/firewall between Exch?

[bart21]

Please provide us with as much info as you can such as the info sukh has asked for.

 

Can you also post the output to

 

Get-databaseavailabilitygroup | fl

 

 

Nick

[Mr.Ben]

Thanks,

 

I seem to have fixed the issue regarding the DAG network, Now a different problem has reared it's head, External Mail isn't getting through (Internal works as expected).

 

I've created their own Network between them (a new VLAN), and assigned the 'Internal DAG' group to a 192.168.0.0/30 range. Both The servers second network card is assigned to the external network (A 10.x.x.x/24 range).

 

A reboot and clear/re-entry of TCPIP settings and the replication is now working.

 

Some more details:

 

There are two servers involved here, one with the Hub Transport, Client Access and Mailbox Role (Call it Email1) and another with just the Mailbox Role (Email2). Each have only 2 network cards, no teaming is enabled.

 

The DAG Group consists of both Email 1 and Email 2

 

The DAG Networks: One on the 192 range (Replication Enabled) and one on the 10 range (Replication Disabled)

 

The 192 Range has no access to the general network, and replication is working across it. The network cards on the 192 range have the Firewall turned off.

 

Email 2 has the active copy of the database.

 

As I understand it, The Transport Server (Email1) should receive the mail, then deliver it the Mailbox Database (Email2) using the external (10.x.x.x) Network

 

Replication of that email (back to Email1) should happen over the private network (192.x.x.x range)

 

*Another note is that for 1400 users there is only 16GB of RAM in each, meaning that they are severely underpowered if you take Microsofts recommendations.

 

So the question is how do I see which network is being used to deliver the mail to the mailboxes?

 

My concern is that the adaptor I have used for the internal DAG network may well have been the adaptor that external email was being delivered to.

 

Which Log can I see the incoming mail on?

Edited September 22, 2012 by Mr.Ben

[EduTech]

Do you use a SmartHost for Inbound/Outbound Email or do you rely on MX Records to deliver the email directly to your Exchange Server? Do you have the relevent ports open i.e. 25 and what type of firewall are you using?

 

If you send an email from an external address > internal address do you get a bounce back from Exchange Server? or/ Do you get an undeliverable message?

 

James.

[Mr.Ben]

Do you use a SmartHost for Inbound/Outbound Email or do you rely on MX Records to deliver the email directly to your Exchange Server? Do you have the relevent ports open i.e. 25 and what type of firewall are you using?

 

If you send an email from an external address > internal address do you get a bounce back from Exchange Server? or/ Do you get an undeliverable message?

 

James.

 

Hi James,

 

Bear with me, I didn't set this up and have no knowledge in my team (we are all less than 5 months in - I don't know what happened!).

 

I don't know on the first part, the SWGfL are a third party in this. The external MX record points to the SWGfL. The non delivery of external mail has happened before and I've had to reboot the server several times and it seemed to magically come back! (However the Replication over the main network issue/multiple entries in DNS etc issue may have been the cause of this - and the fact that it happened again on Friday afternoon when I was correcting the DAG makes me suspicious). I assume that the SWGfL is a Smart Host.

 

External> Internal gives no bounce back messages, the mail just never arrives (this is the really confusing part!). It's not queued either.

 

Thanks again!

Edited September 22, 2012 by Mr.Ben

[EduTech]

Hi James,

 

Bear with me, I didn't set this up and have no knowledge in my team (we are all less than 5 months in - I don't know what happened!).

 

I don't know on the first part, the SWGfL are a third party in this. The external MX record points to the SWGfL. The non delivery of external mail has happened before and I've had to reboot the server several times and it seemed to magically come back! (However the Replication over the main network issue/multiple entries in DNS etc issue may have been the cause of this - and the fact that it happened again on Friday afternoon when I was correcting the DAG makes me suspicious). I assume that the SWGfL is a Smart Host.

 

External> Internal gives no bounce back messages, the mail just never arrives (this is the really confusing part!). It's not queued either.

 

Thanks again!

 

I imagine that your email comes through a Mail Filtering Service provided by SWGfL and I am guessing that the email is being held on that service that is provided by them which I assume you don't have access to (or might but don't know about it yet). I am guessing you have access to OWA so can you telnet on port 25 to your OWA Address? i.e. "telnet mail.yourdomain.com 25" does it show your Exchange Server (Local Name) in the header? This will ensure that your server can accept mail on port 25 coming inbound from your external filtering service.

 

As your not getting a bounce back I assume that is where the mail is going, and if your not seeing the mail on your exchange server then this would make it more so. although, I would have thought the filtering service would have sent a bounceback stating that the mail has not yet being delivered.

 

Can you do me a favour, using ExRCA can you see what that shows if you try sending an email inbound? https://www.testexchangeconnectivity.com/

 

Regards,

James.

[apeman]

I wouldn't worry about the 16GB ram, i have run an exchange 2010 server with nearly 25,000 accounts with only 20GB of ram. Microsofts guidelines are mainly based on concurrent connections, are you going to have all 1400 users connected at the same time??

 

This setup also seems a little strange to me, what is the purpose of the DAG?? and why not just deliver the mail to email1 then replicate it to email2?? instead of passing the email around twice.

[EduTech]

I agree in regards to not worrying about the RAM, It won't be causing the issues you are having at the moment but agree you could potentionally look into it at a later date but don't try and make your infrastructure meet those guidelines. Understand why they say such, but take a lot of it with a pinch of salt

 

I think we could propose a slightly different setup, but lets find out where the hell your mail is going first :-)

 

James.

[Mr.Ben]

The School seems to be a heavy user of Email, 600 PCs on site and generally the RAM on email 2 is maxxed out. When the migration happens back to email1 it couldn't cope and the whole thing has been falling over for some months on a regular basis.

 

Usage on Email one is about 8 GB without the databases active, and on email2 its always 99%.

 

My assumptions on ram usage are based on 500 concurrent connections, so I'm going to add an extra 8GB to start on both email one and two.

 

The purpose of the Dag was (I assume) for backup. I've moved the databases to email2 because of the poor performance of email1 when they are active on it (8GB of ram is taken up by the hub and client access roles alone).

[apeman]

Ok but i still wouldn't be worried with 500 concurrent connections.

Email2 is doing what it it supposed to do i.e. putting as much of the database in ram, if you add more ram you will still see 99% usage as it will just put even more of the database in ram.

 

OK moving the active database to email2 does make sense if email1 only has 16GB of ram also, the cas role can eat memory.

If the thinking behind the dag was to have a real time backup of the database if one of the servers failed and you can deal with the downtime of the failed server then thats fine, but if the person who set it up was thinking it would provide High Availability they are wrong because you still need the other 2 roles for your email systems to still function.

 

 

As for the mail flow problem

 

You have said you cannot receive email from external but can you send internal to internal and internal to external??

As EduTech said can you telnet to your server from outside?? can you still access your OWA??

What changes were you making when the email stopped??

[Mr.Ben]

If I'm honest I don't see why the mailbox role is on email1 except for a disaster recovery scenario rather than a ha solution (I'm virtualising the site this term with ha cluster over this year, so I would move the CAS and HT roles to a virtual server within this - after I've read some more about it!)

 

I can't telnet from the outside world, but the client access externally goes through an ISA server, which would block it - I will add a rule today so that I can.

 

Internal to Internal and Internal to External email is working, and owa is working.

Edited September 23, 2012 by Mr.Ben

[sukh]

So you still got an issue from External>Internal?

 

If yes, what NDR do you get when you send from hotmail to an internal recipient?

 

Have you upped the logging in the receive connector?

[EduTech]

So you still got an issue from External>Internal?

 

If yes, what NDR do you get when you send from hotmail to an internal recipient?

 

Have you upped the logging in the receive connector?

 

He does not get an NDR when sending an email External > Internal, If you do a check on ExRCA it says everything is okay but I believe that it saying everything i okay in the sense it has been delivered to the Email Filtering Service.

 

Attempting to send a test e-mail message to using MX (SWGfL - RM).

 

The test message was delivered successfully.

 

James.

[sukh]

So the mail flow is External>SWGfl>Exch it seems.

If yes, then it would seem that delivery is successful because SWGfl has no issues.

 

To narrow it down, up the logging on the receive connector which accepts email from SWGfl.

 

As a test, if you're 100% sure that email is not being accepted, just check message tracking for a test email from ext>int

[Mr.Ben]

Solved at 10am this morning (Thanks to @EduTech again). The Smarthost was pointing at the DAG Group IP (Which I changed on Friday afternoon, as the DAG group caused more woes than not receiving mail).

 

The adaptor labelled as 'External' connected to the ISA server for Client access only, the receive connectors were pointing at the external address (this confused me, but then I removed the DAG group address on Friday morning (checked my notes!)).

 

I started receiving NDR's saying that the host was not available by 9am this morning.

 

Thanks for all your pointers over the last couple of days!

 

The initial problem of poor owa performance (and the non receiving of external emails) was caused by the DAG group and the send/receive connectors being on the same adaptor.

 

Since 12 I've had no calls about email - first time since I started and it seems the problem is solved, as well as my understanding of exchange 2010 becoming far more advanced!

 

Thanks to all again!

Edited September 24, 2012 by Mr.Ben

[EduTech]

Solved at 10am this morning (Thanks to @EduTech again). The Smarthost was pointing at the DAG Group IP (Which I changed on Friday afternoon, as the DAG group caused more woes than not receiving mail).

 

The adaptor labelled as 'External' connected to the ISA server for Client access only, the receive connectors were pointing at the external address (this confused me, but then I removed the DAG group address on Friday morning (checked my notes!)).

 

I started receiving NDR's saying that the host was not available by 9am this morning.

 

Thanks for all your pointers over the last couple of days!

 

The initial problem of poor owa performance (and the non receiving of external emails) was caused by the DAG group and the send/receive connectors being on the same adaptor.

 

Since 12 I've had no calls about email - first time since I started and it seems the problem is solved, as well as my understanding of exchange 2010 becoming far more advanced!

 

Thanks to all again!

 

Pleased to hear everything is back to normal for you now. Monitor and Ensure any errors are now resolved with Exchange ;-) Event Log should start to be a bit happier with you now that is sorted.

 

Happy Days! Enjoy the rest of your week sir.

 

James.

[synaesthesia]

16gb for Exchange and you're worried? Wow, thank god for Zimbra! For *moments* today I was considering suggesting a move to Exchange!