High Availablility
-
Wednesday, April 15, 2009 8:36 PMI've read about DAG and database mobility, will these be how we refer to CCR/SCR/LCR from now on, or will the existing role names still continue in use?
It's still a little unclear also (you say that LCR will be discontinued), will there be database copies without failover clustering, or will all copies switch on failover clustering?
Finally, backup-less Exchange organisations - are these a good idea? Many of my clients will grasp this concept to reduce cost on backups while improving resilience, but what happens when we need to go back in time to last month?
Sorry for the large number of questions in one post, I think they are all connected.
Thanks
David Lusty
Answers
-
Wednesday, April 15, 2009 8:41 PM
CCR/SCR/LCR are gone, but the technologies that made them all possible are still there (and better!).
Each node can only have 1 copy of a database as all DB names must be unique throughout the forest and if you then had two copies of that DB on a server you would kind of be breaking that rule. Someone please correct me if LCR is still around, but I haven't seen it in the last few months.
If you need to go back in time you can still use log replay lag on the DB copies. You could have 5 copies, 3 up to date, 1 with a 24-hour lag, 1 with a 7-day lag. Within all of that you still have the dumpster and mailbox retention settings. There is also a new admin-only dumpter that can be your secondary safety-net if you want. It's all up to you! :)- Proposed As Answer by Erik Szewczyk Wednesday, April 15, 2009 9:44 PM
- Marked As Answer by Joey MastersonMicrosoft Employee Thursday, April 16, 2009 4:49 AM
All Replies
-
Wednesday, April 15, 2009 8:41 PM
CCR/SCR/LCR are gone, but the technologies that made them all possible are still there (and better!).
Each node can only have 1 copy of a database as all DB names must be unique throughout the forest and if you then had two copies of that DB on a server you would kind of be breaking that rule. Someone please correct me if LCR is still around, but I haven't seen it in the last few months.
If you need to go back in time you can still use log replay lag on the DB copies. You could have 5 copies, 3 up to date, 1 with a 24-hour lag, 1 with a 7-day lag. Within all of that you still have the dumpster and mailbox retention settings. There is also a new admin-only dumpter that can be your secondary safety-net if you want. It's all up to you! :)- Proposed As Answer by Erik Szewczyk Wednesday, April 15, 2009 9:44 PM
- Marked As Answer by Joey MastersonMicrosoft Employee Thursday, April 16, 2009 4:49 AM
-
Wednesday, April 15, 2009 8:52 PMA couple more bits of information:
Failover clustering is required for all of the high availablity options. You can currently lag a copy for up to 14 days.
http://exchangeexchange.com -
Wednesday, April 15, 2009 9:00 PMI think removing LCR ought to be looked at more closely. I can see the technical reasons, but smaller installations with a single server find this feature very compelling (I've not needed it so far, but I'm sure others must have). What would be the single server copy scenario now witout LCR? I do like the simplification around the CCR and SCR though, did you have an answer about the failover cluster requirement? Again this would require clustering which still currently requires Enterprise Server so I assume "standard" copies must still be possible at least between servers?
As I've mentioned in another post, I recently had a client remove the AD user account and mailbox, then purge the mailbox, then remove the database and storage group. They then re-created the user manually and gave them a new mailbox. Although this was a particularly difficult scenario it would have been massively more difficult without the good old tape backup. I'm not saying a well run enterprise system like Microsoft's (must be perfect...right?) wouldn't be fine without a backup, just that I'm not sure it's wise to push having no backups to customers who may not have the same slick system. -
Wednesday, April 15, 2009 9:20 PM
Although LCR is a cool feature I think there were far fewer installs of LCR than Microsoft had expected so it is believed that this won't be a very disruptive omission.
I think many people will agree with you about the "backupless" solution. It definitely requires a very well run IT organization and planning for all possible situations, however it can work. I think for the next few years we will find that most of the SMORGs are going to need to provide some level of backup.
http://exchangeexchange.com -
Wednesday, April 15, 2009 9:21 PM
I think removing LCR ought to be looked at more closely. I can see the technical reasons, but smaller installations with a single server find this feature very compelling (I've not needed it so far, but I'm sure others must have). What would be the single server copy scenario now witout LCR? I do like the simplification around the CCR and SCR though, did you have an answer about the failover cluster requirement? Again this would require clustering which still currently requires Enterprise Server so I assume "standard" copies must still be possible at least between servers?
As I've mentioned in another post, I recently had a client remove the AD user account and mailbox, then purge the mailbox, then remove the database and storage group. They then re-created the user manually and gave them a new mailbox. Although this was a particularly difficult scenario it would have been massively more difficult without the good old tape backup. I'm not saying a well run enterprise system like Microsoft's (must be perfect...right?) wouldn't be fine without a backup, just that I'm not sure it's wise to push having no backups to customers who may not have the same slick system.
IIRC there is no single-server HA option. Perhaps a self-managed snapshot routine?
Yes, any server that is a member of a DAG will need Windows Server 2008 (or R2 when RTM'd) Enterprise Edition as the DAG uses windows failover clustering in the background for a couple tasks. -
Wednesday, April 15, 2009 9:27 PM
Although LCR is a cool feature I think there were far fewer installs of LCR than Microsoft had expected so it is believed that this won't be a very disruptive omission.
I'd possibly hold off on that one, we are a reasonable size Microsoft Partner and Exchange 2007 is only just starting to sell and be rolled out, and I think with the ending of mainstream support for 2003 this week we'll probably see a lot more small companies making the jump since we have been making sure they are aware of that. -
Wednesday, April 15, 2009 9:28 PM
IIRC there is no single-server HA option. Perhaps a self-managed snapshot routine?
Yes, any server that is a member of a DAG will need Windows Server 2008 (or R2 when RTM'd) Enterprise Edition as the DAG uses windows failover clustering in the background for a couple tasks.
so for our smaller clients what is the recommendation? I assume Linux and OSS isn't the official line? ;o) -
Wednesday, April 15, 2009 9:31 PM
so for our smaller clients what is the recommendation? I assume Linux and OSS isn't the official line? ;o)
I'm not MSFT, so I couldn't say. :) If I had smaller clients I'd probably try to sell them two cheaper servers because now you have role coexistence (Clustered MBX can have HUB & CAS on it now too), which you did not have in 2007 and it makes an over HA solution pretty cheap since 2007 required at least 3 severs before. Or if it is a good fit, run two virtual servers on one slightly more powerful server within a MSFT approved hypervisor.
Remember... one Windows Server Enterprise license = 5 licenses in total. 1 physical & 4 virtual. -
Wednesday, April 15, 2009 9:38 PM
I like the idea, but now instead of spending around £300 on a single Server Std licence, we'll need at least a single (assuming virtualization) Server Enterprise Licence for around £1300. We'll also need 2 Exchange Std licences instead of 1 since that is per instance licensing, as well as better hardware to cope with running 3 instances of Windows, 2 Exchange and the DAG overhead.
I realise that a lot of people are getting excited by "cloud" technology, but not everyone is happy with offsite data.
Thanks for your replies, I don't expect you to have all the answers I'm just making sure these things are considered before some incredibly useful features are removed. -
Wednesday, April 15, 2009 9:48 PMOne of the improvements over 2007 is that you can now have a fully redundant two-server implementation. You can install the Client Access and Hub Transport roles on an HA Mailbox role. This would have been a minimum of three servers prior.
Microsoft Exchange MVP | http://exchangeexchange.com -
Wednesday, April 15, 2009 9:51 PM
One of the improvements over 2007 is that you can now have a fully redundant two-server implementation. You can install the Client Access and Hub Transport roles on an HA Mailbox role. This would have been a minimum of three servers prior.
Microsoft Exchange MVP | http://exchangeexchange.com
am I imagining that you could have a redundant database on a single server implementation in 2007?
Also, a separate question but related - on E2007 a mailbox server would not use a HT server other than itself if it was an HT server. Has this been fixed? if not fully redundant is going to be an issue with 2 servers :o) -
Wednesday, April 15, 2009 9:54 PM
If you do LCR you could have redundant storage, and that could be done with HT and CAS so in theory you could have redundancy throughout. However the key is that you dont get automated failover of services.One of the improvements over 2007 is that you can now have a fully redundant two-server implementation. You can install the Client Access and Hub Transport roles on an HA Mailbox role. This would have been a minimum of three servers prior.
Microsoft Exchange MVP | http://exchangeexchange.com
am I imagining that you could have a redundant database on a single server implementation in 2007?
Also, a separate question but related - on E2007 a mailbox server would not use a HT server other than itself if it was an HT server. Has this been fixed? if not fully redundant is going to be an issue with 2 servers :o)
With Exchange 2010 you can have full redundancy with automatic failover of services within a 2 server install. The caveot here is that you would need to rely on something external to provide load balancing for the CAS role (i.e. ISA, hardware load balancers, etc.)
Erik
Microsoft Certified Master: Exchange 2007; MCSE:M 2003; MCITP: 2008; Windows Server MVP; CCNA -
Wednesday, April 15, 2009 9:56 PM
LCR is gone though?If you do LCR you could have redundant storage, and that could be done with HT and CAS so in theory you could have redundancy throughout. However the key is that you dont get automated failover of services.
-
Wednesday, April 15, 2009 10:02 PM
LCR is gone though?If you do LCR you could have redundant storage, and that could be done with HT and CAS so in theory you could have redundancy throughout. However the key is that you dont get automated failover of services.
Correct LCR is gone, I was responding specifically to the comment about enabling full redundancy on Exchange 2007 with 2 servers.
Microsoft Certified Master: Exchange 2007; MCSE:M 2003; MCITP: 2008; Windows Server MVP; CCNA -
Wednesday, April 15, 2009 10:06 PM
With Exchange 2010 you can have full redundancy with automatic failover of services within a 2 server install. The caveot here is that you would need to rely on something external to provide load balancing for the CAS role (i.e. ISA, hardware load balancers, etc.)
Erik, I forget Is WNLB not supported on the server if MBX2010 or HUB2010 also coexists? I know you can use WNLB for CAS2010 only servers. -
Wednesday, April 15, 2009 10:13 PM
Erik, I forget Is WNLB not supported on the server if MBX2010 or HUB2010 also coexists? I know you can use WNLB for CAS2010 only servers.
If that is the case then availablility would actually be reduced due to the failover time of clustering compared to simply using another server.
I agree these new features are a vast improvement over 2007, but removing some features has removed a few deployment options as well which will adversly affect quite a few customers.
I realise I'm sounding a little pedantic on this one, but I design an Exchange system roughly once/twice a week so I need to understand all of the deployment scenarios and overall it looks like we'll be worse off to me. -
Wednesday, April 15, 2009 10:20 PM
You can use WNLB with CAS for sure (which is part of Server Standard) and I don't think it harms HUB either since mailbox servers point to HUB by name. I just forget at this moment if WNLB messes with the mailbox role or not. I don't think so, but someone else may know for sure.If that is the case then availablility would actually be reduced due to the failover time of clustering compared to simply using another server.
I agree these new features are a vast improvement over 2007, but removing some features has removed a few deployment options as well which will adversly affect quite a few customers.
I realise I'm sounding a little pedantic on this one, but I design an Exchange system roughly once/twice a week so I need to understand all of the deployment scenarios and overall it looks like we'll be worse off to me. -
Wednesday, April 15, 2009 10:53 PM
Erik, I forget Is WNLB not supported on the server if MBX2010 or HUB2010 also coexists? I know you can use WNLB for CAS2010 only servers.
If that is the case then availablility would actually be reduced due to the failover time of clustering compared to simply using another server.
I agree these new features are a vast improvement over 2007, but removing some features has removed a few deployment options as well which will adversly affect quite a few customers.
I realise I'm sounding a little pedantic on this one, but I design an Exchange system roughly once/twice a week so I need to understand all of the deployment scenarios and overall it looks like we'll be worse off to me.Hi Dave,
I ask you give failover a go in your lab, I think you will be surprised at the failover times between databases/servers in a DAG, I would be surprised if users in OWA would even notice, in fact I have not heard of any reports that a "working" database failover effected OWA users, at the moment the failover story for Outlook is not to good as it asks you to close and re-open Outlook, but I would expect that to improve at RTM.
LCR has gone, however if you are dealing with a standard standalone server for a medium size org, you do have the option of using VSS based snapshot backup solution, which one could presume would have been deployed to backup other services and applications.
And you don't sound pedantic, it is alot to get your head round, but please give it ago it is worth it. -
Wednesday, April 15, 2009 10:55 PM
With Exchange 2010 you can have full redundancy with automatic failover of services within a 2 server install. The caveot here is that you would need to rely on something external to provide load balancing for the CAS role (i.e. ISA, hardware load balancers, etc.)
Erik, I forget Is WNLB not supported on the server if MBX2010 or HUB2010 also coexists? I know you can use WNLB for CAS2010 only servers.
WNLB will work with any of the Exchange roles, however it does not play nice with a WFC. Which means if you want 2 boxes with HA for all roles you cannot use WNLB.
Erik
Microsoft Certified Master: Exchange 2007; MCSE:M 2003; MCITP: 2008; Windows Server MVP; CCNA -
Wednesday, April 15, 2009 11:00 PMToo many discussion topics in this thread so hard to keep track of everything :)
Failover times: OWA still has some issues failing over, but from what I've seen generally nothing that a refresh cannot fix.
Outlook restart: If Outlook is prompting you to restart it means that your outlook client is connecting directly to the mailbox server rather than a Client Access Array (MoMT) running on your CAS. If you configure a Client Access Array then all client MAPI connections terminate on the CAS and they obscure the failover and you wont have to restart the client.
Erik
Microsoft Certified Master: Exchange 2007; MCSE:M 2003; MCITP: 2008; Windows Server MVP; CCNA -
Thursday, April 16, 2009 6:14 AM
WNLB will work with any of the Exchange roles, however it does not play nice with a WFC. Which means if you want 2 boxes with HA for all roles you cannot use WNLB.
Erik
Microsoft Certified Master: Exchange 2007; MCSE:M 2003; MCITP: 2008; Windows Server MVP; CCNA
in my experience, WNLB is always available, so for CAS this is a requirement for HA. HT is happy to use more than one server anyway as long as it's not on the MBX server so doesn't require WNLB or WFC.
WFC failovers are far from instantaneous on production servers since the entire identity of the server along with networking needs to be moved accross to the passive node. This includes kerberos registration etc.
Again, I'm not saying I don't love the new functionality, but I'm a little miffed at losing so many very useful features, and seemingly only because they were not perceived to be used very much. -
Thursday, April 16, 2009 3:21 PM
Outlook restart: If Outlook is prompting you to restart it means that your outlook client is connecting directly to the mailbox server rather than a Client Access Array (MoMT) running on your CAS. If you configure a Client Access Array then all client MAPI connections terminate on the CAS and they obscure the failover and you wont have to restart the client.
Not to sound dense, but what is a 'Client Access Array (MoMT)'? Is this something new to Exchange 2010 or are you talking about Outlook clients running in cached mode (Outlook Anywhere)? I would love to have my non-cached mode clients connect to the CAS instead of connecting directly to the mailbox servers. -
Thursday, April 16, 2009 3:35 PM
Outlook restart: If Outlook is prompting you to restart it means that your outlook client is connecting directly to the mailbox server rather than a Client Access Array (MoMT) running on your CAS. If you configure a Client Access Array then all client MAPI connections terminate on the CAS and they obscure the failover and you wont have to restart the client.
Not to sound dense, but what is a 'Client Access Array (MoMT)'? Is this something new to Exchange 2010 or are you talking about Outlook clients running in cached mode? I would love to have my non-cached mode clients connect to the CAS instead of connecting directly to the mailbox servers.
In response to this and Dave Lustly's comments regarding WFC reconnect times.
The concept of the Clustered Mailbox Server is gone, in addition the MAPI endpoint for clients is being moved to the CAS (which proxies back to the appropreate mailbox server). With clients connecting to CAS the failover of a mailbox database between nodes becomes much more transparent because you maintain connectivity to the CAS, there is still some disruption as the database is mounted however it is reduced. Both of these are new concepts to Exchange 2010.
I'm not really finding a good doc for it yet, probably has yet to be written. These docs do touch on these concepts but unfortunaly not in great depth:
http://technet.microsoft.com/en-us/library/dd346700(EXCHG.140).aspx
http://technet.microsoft.com/en-us/library/dd335211(EXCHG.140).aspx
That's part of the fun thing about playing with a Beta, not all the details are available yet :)
Erik
Microsoft Certified Master: Exchange 2007; MCSE:M 2003; MCITP: 2008; Windows Server MVP; CCNA- Proposed As Answer by Erik Szewczyk Thursday, April 16, 2009 5:00 PM
-
Thursday, April 16, 2009 4:57 PM See also
http://technet.microsoft.com/en-us/library/dd633496(EXCHG.140).aspx
http://technet.microsoft.com/en-us/library/dd351223(EXCHG.140).aspx
http://technet.microsoft.com/en-us/library/dd298065(EXCHG.140).aspx
http://technet.microsoft.com/en-us/library/dd335158(EXCHG.140).aspx
Hope this helps.
This posting is provided "AS IS" with no warranties, and confers no rights.- Proposed As Answer by Scott Schnoll [MSFT] Thursday, April 16, 2009 4:58 PM
-
Thursday, April 16, 2009 5:00 PM
See also
http://technet.microsoft.com/en-us/library/dd633496(EXCHG.140).aspx
http://technet.microsoft.com/en-us/library/dd351223(EXCHG.140).aspx
http://technet.microsoft.com/en-us/library/dd298065(EXCHG.140).aspx
http://technet.microsoft.com/en-us/library/dd335158(EXCHG.140).aspx
Hope this helps.
This posting is provided "AS IS" with no warranties, and confers no rights.
Thanks Scott!
Microsoft Certified Master: Exchange 2007; MCSE:M 2003; MCITP: 2008; Windows Server MVP; CCNA -
Friday, April 17, 2009 6:26 AM
I think my point on this one was that if the CAS is also on the cluster server then it will have to fail over as well, so you lose that exact transparency you're speaking of.
In response to this and Dave Lustly's comments regarding WFC reconnect times.
The concept of the Clustered Mailbox Server is gone, in addition the MAPI endpoint for clients is being moved to the CAS (which proxies back to the appropreate mailbox server). With clients connecting to CAS the failover of a mailbox database between nodes becomes much more transparent because you maintain connectivity to the CAS, there is still some disruption as the database is mounted however it is reduced. Both of these are new concepts to Exchange 2010.
I do understand the new clustering, but under the hood it's still a failover node hosting the mailbox database which has a server name and IP address, you just have to translate the marketing speak in the documents. -
Friday, April 17, 2009 6:26 AM
thanks for those, I'll have a readSee also
http://technet.microsoft.com/en-us/library/dd633496(EXCHG.140).aspx
http://technet.microsoft.com/en-us/library/dd351223(EXCHG.140).aspx
http://technet.microsoft.com/en-us/library/dd298065(EXCHG.140).aspx
http://technet.microsoft.com/en-us/library/dd335158(EXCHG.140).aspx
Hope this helps.
This posting is provided "AS IS" with no warranties, and confers no rights.
Thanks Scott!
Microsoft Certified Master: Exchange 2007; MCSE:M 2003; MCITP: 2008; Windows Server MVP; CCNA -
Friday, April 17, 2009 4:55 PM
I've found two distinct information concerning HA:
"Multiple server roles can co-exist on servers that provide high availability. This enables small organizations to deploy a two-server configuration provides full redundancy of mailbox data, while also providing redundant Client Access and Hub Transport services."
From article "New High Availability Functionality"
"The server hosting the file share witness and directory cannot be a member of the DAG that is using that share and directory. We recommend that you use a Hub Transport server to host the file share witness, as this allows an Exchange administrator to be aware of the availability of the file share witness."
From article "Create a Database Availability Group"
The latter clearly states that FSW must be in different server (Hub) than in DAG member server (that is Mailbox Role). There's a little contradiction between those two information. How many server's you really need to have to provide full reduntant Exchange enviroment without 3rd party solutions? I guess four as in Exchange 2007.
Maybe I haven't understood the concept of HA correctly yet? -
Friday, April 17, 2009 6:38 PMPetri:
You'll need three machines -- two Exchange servers and a third machine for the FSW. The third machine does not need an Exchange licence.
In a larger deployment it's recommended that the FSW is on an Exchange Server. When running add-databaseavailabilitygroupserver, the task will create the file share on that other computer. If it's Exchange Server, then the firewall exceptions should have been pre-installed, and the account running the task would have the rights to create the share. (Actually, I don't remember if the Beta has the firewall configuration code, or if that went in after.)
When the FSW is on a non-Exchange machine, then the firewall may not be configured correctly, and the task might get ACCESS_DENIED when creating the network share.
If you create the share manually, it needs to grant the DAG computer account Full Control.
If you only have the two machines of the DAG, then when one machine goes down, the other will lose quorum, and you'll lose availability. The FSW prevents that downtime.
-martin
Legal Stuff: This posting is provided "AS IS" with no warranties, and confers no rights. -
Friday, April 17, 2009 6:57 PM
Petri:
Thank you for your answer.
You'll need three machines -- two Exchange servers and a third machine for the FSW. The third machine does not need an Exchange licence.
In a larger deployment it's recommended that the FSW is on an Exchange Server. When running add-databaseavailabilitygroupserver, the task will create the file share on that other computer. If it's Exchange Server, then the firewall exceptions should have been pre-installed, and the account running the task would have the rights to create the share. (Actually, I don't remember if the Beta has the firewall configuration code, or if that went in after.)
When the FSW is on a non-Exchange machine, then the firewall may not be configured correctly, and the task might get ACCESS_DENIED when creating the network share.
If you create the share manually, it needs to grant the DAG computer account Full Control.
If you only have the two machines of the DAG, then when one machine goes down, the other will lose quorum, and you'll lose availability. The FSW prevents that downtime.
-martin
Legal Stuff: This posting is provided "AS IS" with no warranties, and confers no rights.
-Petri -
Friday, April 17, 2009 8:20 PM
I think my point on this one was that if the CAS is also on the cluster server then it will have to fail over as well, so you lose that exact transparency you're speaking of.
In response to this and Dave Lustly's comments regarding WFC reconnect times.
The concept of the Clustered Mailbox Server is gone, in addition the MAPI endpoint for clients is being moved to the CAS (which proxies back to the appropreate mailbox server). With clients connecting to CAS the failover of a mailbox database between nodes becomes much more transparent because you maintain connectivity to the CAS, there is still some disruption as the database is mounted however it is reduced. Both of these are new concepts to Exchange 2010.
I do understand the new clustering, but under the hood it's still a failover node hosting the mailbox database which has a server name and IP address, you just have to translate the marketing speak in the documents.
No, it's not a failover node that's hosting the mailbox database. There is a named cluster and and IP address, however it could be brought completely offline and Exchange would continue to serve client requests. WFC is used under the hood to maintain quorum and heartbeat, but it's Exchange that is managing the availability of the mailbox databases.
As for hosting CAS on the mailbox server you're right in that if the server completly goes down the client will have to recconect to a CAS. However this is provided via a load balancer that is not state aware, so the failover complexity/time requirements go down signifigantly.
Microsoft Certified Master: Exchange 2007; MCSE:M 2003; MCITP: 2008; Windows Server MVP; CCNA- Proposed As Answer by Erik Szewczyk Friday, April 17, 2009 8:20 PM
-
Thursday, April 23, 2009 2:21 PMIt's also not really clear for me.
I installed 1 windows 2008 R2 standard for DC and 2 windows 2008 R2 enterprise with exchange 2010 beta. Each node hosts following roles (HUB, CAS, MAILBOX)
I created a DAG named dag1.beta.local with FSW hosted on DC.
Active database was on the first Exchange node.
When connected with OWA (https://dag1.beta.local), i can move the active mailbox database to the second node. OWA still works.
But with outlook 2007 SP1 in MAPI (cached or not), once active mailbox moved to the second node, outlook never reconnects on the mailbox.
Each exchange server has 2 network cards on different subnet. Replication is enabled only on the second subnet.
Another thing, i entered dag1.beta.local in exchange server field. when i click on "Check Name" button, outlook changes exchange server name always to the 1st exchange node (mailbox database active or not).
any idea ?
regards
Stéphane -
Thursday, April 23, 2009 2:27 PM
Outlook 2007 Sp1 does not have the capability to connect automatically. You have to close and open outlook.
-
Thursday, April 23, 2009 3:09 PMI tried to close and open outlook again. Unable to connect to mailbox until active mailbox database has been moved back to the 1st node.
-
Monday, June 08, 2009 3:44 PMOK let me see if I have this right for Exchange 2010:
We have 3 reginoal offices and DR site I would need to take in consideration. I am not worried about Databases in DAG, I think I have grip on that.
A big change in exchange 2010 is the client no longer coonects to the Database server but the CAS. So we need have HA in CAS.
Since I have total of 5 locations I need to worry about I will need 2 CAS servers in each with either Cisco ACE or Citrix NetScaler is fronting those servers.
Those server though can also be the DB servers. Correct?
Also the 1st client that can auto sense the CAS change is Outlook 2007 SP2 Correct? -
Wednesday, June 17, 2009 9:00 AMI have 1 server 2K8DC (Hub, CAS,DC) , 2 mailbox server role EX1, EX2. EX1 and EX2 was installed Failover clustering feature by manually. All of them were installed in VMWARE 6.5 (windows 2008 Enterprise SP2)
I used Local Poweshell at EX1 to run command
New-DatabaseAvailabilityGroup -Name DAG1 -FileShareWitnessShare \\2K8DC\DAG1 -FileShareWitnessDirectory C:\DAG1 (success but there is no folder DAG1 in DC)
Add-DatabaseAvailabilityGroupServer -Identity DAG1 -MailboxServer EX1 -DatabaseAvailabilityGroupIpAddresses 192.168.1.50 (success)
Add-DatabaseAvailabilityGroupServer -Identity DAG1 -MailboxServer EX2 (failed)
WARNING: The operation wasn't successful because an error was encountered. More
details may be found in log file
"C:\Users\Administrator\AppData\Local\Temp\dagtask_2009-06-17_08-53-25_add-data
baseavailabiltygroupserver.log".
Add-DatabaseAvailabilityGroupServer : Cluster common transient exception: AddCl
usterNode failed with -2147024891, MaxPercentage=12.
At line:1 char:36
+ Add-DatabaseAvailabilityGroupServer <<<< -Identity DAG1 -MailboxServer EX2
+ CategoryInfo : InvalidArgument: (:) [Add-DatabaseAvailabilityGr
oupServer], ClusCommonTransientException
+ FullyQualifiedErrorId : 85A07DFE,Microsoft.Exchange.Management.SystemCon
figurationTasks.AddDatabaseAvailabilityGroupServer
I checked the log file
WriteError! Exception = Microsoft.Exchange.Cluster.Replay.ClusCommonTransientException: Cluster common transient exception: AddClusterNode failed with -2147024891, MaxPercentage=12. ---> System.ComponentModel.Win32Exception: AddClusterNode failed.
--- End of inner exception stack trace ---
I re-installed Failover clustering feature on EX2, it's not work -
Tuesday, June 23, 2009 7:01 PMsuthuc, that's ERROR_ACCESS_DENIED. There are some more suggestions in the thread 'Creating a DAG'.
In the Beta build you need to run Add-DatabaseAvailabilityGroupServer from an elevated local powershell. Remote powershell will give you the ERROR_ACCESS_DENIED.
-martin
Legal Stuff: This posting is provided "AS IS" with no warranties, and confers no rights. -
Wednesday, June 24, 2009 2:16 AMhi Martin,
You mean I need to run Add-DatabaseAvailabilityGroupServer -Identity DAG1 -MailboxServer EX1 -DatabaseAvailabilityGroupIpAddresses 192.168.1.50 with Local Power Shell at EX1
and run Add-DatabaseAvailabilityGroupServer -Identity DAG1 -MailboxServer EX2 with Local Power Shell at EX2 -
Wednesday, June 24, 2009 2:21 AM
It shouldn't matter where you run it. The important part is that it's Local PowerShell.
When you run local powershell, the operations get run under your account.
The remote powershell runs the operations in w3wp.exe, which is running as LOCALSYSTEM (the machine account). The operations get sent to w3wp.exe, executed there, and the results are sent back. The machine account is an entirely different account that has different permissions (and a slightly different authentication model). The machine account's different permissions is why it ends up with the 'Access Denied' error.
I hope that made sense.
-martin
Legal Stuff: This posting is provided "AS IS" with no warranties, and confers no rights. -
Wednesday, June 24, 2009 2:45 AMThat's right, Martin
I've ever run both command at EX1, the error occurs. It's the same error if I ran those command at EX2. I have not tried your guidance yet, so not know the results yet. May you suggest what else? -
Thursday, July 16, 2009 6:13 PMI can use DAG in Hyper-V environment. But when I remove DAG and try again, it's not working. ......"watson dump"..... That's is Exchange 2010 beta. It does not work smoothly
-
Wednesday, July 29, 2009 1:58 AMThis is my lab step by step DAG in Exchange server 2010. http://thuc.nhatnghe.vn/ex2010-dag
-
Wednesday, September 30, 2009 8:47 PMIf I understand it correctly, MS has simplified the process of bringing up database copies online in 2010 by integrating all the functionality within EMC. In 2007 you have to run a series of powershell scripts in order to make an SCR database mountable.
However I don't see any easy way now of re-homing Outlook clients to the server with the new active copy. It seems you are now forced to use at least ONE dedicated CAS server, correct? Unless of course you want to instruct all of your users to manually change their Outlook profiles.
In Exchange 2007 you could run the following powershell script to re-home mailboxes and Outlook would pick up the change:
a. Get-Mailbox -Database "server1\first storage group\mailbox database" |where {$_.ObjectClass –NotMatch '(SystemAttendantMailbox|ExOleDbSystemMailbox)'}| Move-Mailbox -ConfigurationOnly -TargetDatabase "server2\first storage group\mailbox database”
However, this no longer works under Exchange 2010.
So back to the original, slightly modified question: Assuming I don't want to use multiple dedicated CAS servers (isn't that going to now approximately double my licensing costs?), how do I re-home mailboxes to the new server so Outlook autodiscover will know where to go? -
Tuesday, November 10, 2009 8:33 PMCan anyone expound on this? It's great that database failover is simpler now but what about getting the clients to talk to the new database server without a dedicated CAS machine or having to manually change their Outlook profile?
.png){kind=spacer}
