Tim McMichael

Navigating the world of high availability...and occasionally sticking my head in the cloud...

Exchange 2010 – Stopping the Information Store Service does not failover database copies.

Exchange 2010 – Stopping the Information Store Service does not failover database copies.

  • Comments 5
  • Likes

In Exchange 2010 we achieve high availability of mailbox databases by utilizing a Database Availability Group (DAG).  When a DAG is utilized, and mailbox database copies are created on DAG members, they are either MOUNTED (active) or have a copy status – for example HEALTHY (passive).

If an administrator or another process gracefully stops the Information Store Service on a DAG node, any database copies that were mounted on that server enter a DISMOUNTED state.  These copies do not fail over to another node, as the shutdown was graceful and not the result of an error condition that would trigger a failover event to occur.  Should the processed have crashed or otherwise became unavailable, this would be detected as an error and the database instances failed over.

Let’s take a look at this.

In this environment I have a four mailbox server DAG.  Database DAG-DB0 is replicated to all members of the DAG.  Currently DAG-DB0 is mounted on mailbox server DAG-1.

Get-mailboxdatabasecopystatus *\DAG-1

Name                                          Status        
----                                          ------         
DAG-DB0\DAG-1                                 Mounted             
DAG-1-DB0\DAG-1                               Mounted          
DAG-DB1\DAG-1                                 Healthy
 

On DAG-1 I issued a net stop msexchangeis.  This gracefully stopped the Information Store Service.

Log Name:      System
Source:        Service Control Manager
Date:          3/21/2010 11:27:19 AM
Event ID:      7036
Task Category: None
Level:         Information
Keywords:      Classic
User:          N/A
Computer:      DAG-1.domain.com
Description:
The Microsoft Exchange Information Store service entered the stopped state.

Reviewing the mailbox database copy status for DAG-1, the following is noted.  (Get-mailboxdatabasecopystatus *\DAG-1)

Name                                          Status         
----                                          ------         
DAG-DB0\DAG-1                                 Dismounted     
DAG-1-DB0\DAG-1                               Dismounted     
DAG-DB1\DAG-1                                 Healthy
 

As a response to the IS gracefully shutting down the mailbox database copies that were mounted on that host are now dismounted.

Subsequently I issued the command net start msexchangeis to start the Information Store Service. 

Log Name:      System
Source:        Service Control Manager
Date:          3/21/2010 11:34:18 AM
Event ID:      7036
Task Category: None
Level:         Information
Keywords:      Classic
User:          N/A
Computer:      DAG-1.domain.com
Description:
The Microsoft Exchange Information Store service entered the running state.

After restarting the service the databases that were previous dismounted are now mounted.  (Get-mailboxdatabasecopystatus *\DAG-1)

Name                                          Status         
----                                          ------         
DAG-DB0\DAG-1                                 Mounted        
DAG-1-DB0\DAG-1                               Mounted        
DAG-DB1\DAG-1                                 Healthy

Comments
  • Good post Tim. Just had this happen....NetApp service account stopped the IS and the database didnt fail over to the local copy.

  • Do you think that if I wanted to restart the Exchange server and first stopped the IS, that the DAG would not fail over with the restart? We know that it will if we just restart the box, which to me seemed like a "graceful" shutdown of the IS. We are trying to come up with the procedure to restart an Exchange server in a DAG and keep the databases from failing over.

  • @Michael Commons:

    I too first thought that stopping a DAG node was considered a graceful IS shutdown - it is not.

    When the DAG node stops and the cluster service stops, other DAG nodes detect this action and will begin to move database instances to other servers that remain online.

    I believe your only optoin would be to issue a dismount-database prior to restarting the machine.  This will also set an active manager flag indicating the database was admin dismounted.  When the node reboots, the "active" copy will still be arbitrated to another machine, just not arbitrated in a mounted state.

    After the member is backup, you'd have to move the instances back and then mount them.

  • I need to restart the information service on the mailbox server which is hosting the active copies of our databases in our 2 member DAG. I need to do this after entering the Exchange product license key. Based on your results it looks like it do not need to move the active database copies before restarting the service - is my understanding correct?

  • @John:

    Apologize for my delay in responding - I was offline.

    You shouls always desire to move your database copies to the passive node for maintenance like this.

    TIMMCMIC

Your comment has been posted.   Close
Thank you, your comment requires moderation so it may take a while to appear.   Close
Leave a Comment