• RSS
  • Twitter
  • FaceBook

Exchange Server Forums

Forums | Register | Login | My Profile | Inbox | RSS RSS icon | My Subscription | My Forums | Address Book | Member List | Search | FAQ | Ticket List | Log Out

Disable Database dismount when WFS lost

Users viewing this topic: none

Logged in as: Guest
  Printable Version
All Forums >> [Microsoft Exchange 2010] >> High Availability >> Disable Database dismount when WFS lost Page: [1]
Login
Message << Older Topic   Newer Topic >>
Disable Database dismount when WFS lost - 4.Nov.2011 12:59:08 PM   
InfMail

 

Posts: 5
Joined: 20.May2011
Status: offline
Is there anyway to stop an active database dismounting when Quorum is lost.

The scenario is : two sites, each site has their own Exchange seever (all roles) that have an active database for local users and a copy of the other site's mailbox database for contingency. The passive database is prevented from automatically mounting.

Is it possible to use the alternate WFS of the other site to avoid.

We're looking for a site resilient, switch over solution rather than HA failover.
Post #: 1
RE: Disable Database dismount when WFS lost - 6.Nov.2011 5:45:48 AM   
ermanishchawla

 

Posts: 125
Joined: 12.Jul.2011
Status: offline
Assuming that you are running two member's DAG. Lets see what you are trying to say

Site A: DAG Member M1
Site A: Witness Server W1
Site B: DAG Member M2
Site C: Alternate Witness Server W2

Please understand the concept that AWS is not a redundancy for WS ie when WS is unavailable DAG will not switch over to AWS to attain a quorum

Now lets see how quoroms are built and broken

1. M1 placed a SMB lock on W1 and hence got the quorom and can serve the users
2. M2 Can connect to locking node and hence all attains a quorom and can serve as well

Now if WAN link is disturbed between Site 1 & Site 2 M1 Holds a quorom since W1 is in same site but M2 wont be able to contact locking node and hence cannot attain a quorum so dismount it.

Now in your case you are trying to use witness server located in other site by change dag properties and make AWS as witness Server in that case due to WAN outage Now Site 1 will loose quorum. So there would be no success at all.

This simple rule explain you everything

The reason for this behavior has to do with the core rules around quorum and DAGs, specifically:
  • All DAGs and DAG members require quorum to operate. If you donít have quorum, you donít have an operational DAG. When quorum is lost, databases are dismounted, connectivity is unavailable and replication is stopped.
  • Quorum requires a majority of voters to achieve a consensus.Thus, when you have an even number of members in a DAG, you need an external component to provide a weighted vote for one of the actual quorum voters to prevent ties from occurring.
    • In a Windows Failover Cluster, only members of the cluster are quorum voters. When the cluster is one vote away from losing quorum and the Witness Server is needed to maintain quorum, one of the DAG members that can communicate with the Witness Server places a Server Message Block (SMB) lock on a file called witness.log that is located in the Witness Directory. The DAG member that places the SMB lock on this file is referred to as the locking node. Once an SMB lock is placed on the file, no other DAG member can lock the file.
    • The locking node then acquires a weighted vote; that is, instead of its vote counting for 1, it counts for 2 (itself and the Witness Server).
    • If the number of members that can communicate with the locking node constitutes a majority, then the members in communication with the locking node will maintain quorum and continuing servicing clients. DAG members that cannot communicate with the locking node are in the minority, and they lose quorum and terminate cluster and DAG operations.

  • The majority formula for maintaining quorum is V/2 + 1, the result of which is always a whole number. The formula is the number of voters (V) divided by 2, plus 1 (for tie-breaking purposes).


< Message edited by ermanishchawla -- 6.Nov.2011 6:04:30 AM >

(in reply to InfMail)
Post #: 2
RE: Disable Database dismount when WFS lost - 7.Nov.2011 5:41:38 AM   
InfMail

 

Posts: 5
Joined: 20.May2011
Status: offline
Thank you very much for the indepth explanation regarding the WFS and AWS. So AWS is not an option.

Our desire is to stop the remote exchange server from dismounting it's active mailbox database (MB2) if it loose Quorum through an outage between sites. The copied mailbox database (MB2) held at the local site is disabled from automounting; so, the chance of split brain is minimal.

Is there any way to disable the automatic dismount of the active mailbox database at the remote site when quorum is lost.

We don't have the budget for a four server 2 DAG configuration. Site resilience (switchover) is the only requirement, not HA.

(in reply to ermanishchawla)
Post #: 3
RE: Disable Database dismount when WFS lost - 7.Nov.2011 8:48:34 AM   
ermanishchawla

 

Posts: 125
Joined: 12.Jul.2011
Status: offline
forcing a quorum might help try with that and let us know outputs and difficulties

< Message edited by ermanishchawla -- 7.Nov.2011 8:51:50 AM >

(in reply to InfMail)
Post #: 4
RE: Disable Database dismount when WFS lost - 8.Nov.2011 4:08:46 AM   
InfMail

 

Posts: 5
Joined: 20.May2011
Status: offline
Isn't forcing a quroum an option only available after the connectivity is lost, when trying to mount the database that has been dismounted ?

(in reply to ermanishchawla)
Post #: 5
RE: Disable Database dismount when WFS lost - 10.Nov.2011 11:50:58 AM   
ermanishchawla

 

Posts: 125
Joined: 12.Jul.2011
Status: offline
Yes it is, it is manual operation only but can be scripted using powershell to automate on triggers  

(in reply to InfMail)
Post #: 6
RE: Disable Database dismount when WFS lost - 10.Nov.2011 12:31:52 PM   
InfMail

 

Posts: 5
Joined: 20.May2011
Status: offline
My understanding is that, in our two server stretched DAG scenario, to force a quorum requires the disabling of the exchange DAG and clustering. To accomodate an occasional outage between sites this is overkill. It appears the only viable solution is 2 sites, 2 DAGS, 4 servers. Costly SR.

Thanks for the response.

(in reply to ermanishchawla)
Post #: 7
RE: Disable Database dismount when WFS lost - 16.Nov.2011 11:04:21 AM   
Jesper Bernle

 

Posts: 912
Joined: 15.Oct.2007
Status: offline
quote:

ORIGINAL: InfMail

My understanding is that, in our two server stretched DAG scenario, to force a quorum requires the disabling of the exchange DAG and clustering. To accomodate an occasional outage between sites this is overkill. It appears the only viable solution is 2 sites, 2 DAGS, 4 servers. Costly SR.

Thanks for the response.


If you (like in your environment) go for the Active/Active model, 2 DAGs is the only way as you concluded yourself. And this is a good thing. If you could 'disable' the dismount you'd run into the dreaded Split Brain Syndrome which ultimatly makes you loose data.

The Exchange Team has for this reason also developed even further safe guarding with the introduction of DAGs in DAC mode.

Read more about this feature here: Understanding Datacenter Activation Coordination Mode

(in reply to InfMail)
Post #: 8

Page:   [1] << Older Topic    Newer Topic >>
All Forums >> [Microsoft Exchange 2010] >> High Availability >> Disable Database dismount when WFS lost Page: [1]
Jump to:

New Messages No New Messages
Hot Topic w/ New Messages Hot Topic w/o New Messages
Locked w/ New Messages Locked w/o New Messages
 Post New Thread
 Reply to Message
 Post New Poll
 Submit Vote
 Delete My Own Post
 Delete My Own Thread
 Rate Posts


Follow TechGenix on Twitter