November 20, 2017 at 5:05 am
Hi Experts,
We have a 2 Node Active \Active Cluster of BizTalk and 2 Node Active\Passive cluster for Application. The Second node for both Application and BizTalk SQL Server resides in Data Center 2 . If the network connectivity between the Data Centers are lost and connection to primary data center is no more available ,what will happen to SQL Server. Will the SQL Server resources in Node 1 which is in Data center 1 will migrate to Node 2?
Please Help
November 20, 2017 at 5:17 am
It will depends on the quorum options you have defined.
You need a witness so disk or fileshare usually and that will determine what would happen if you get failover or shutdown of the services.
November 20, 2017 at 7:54 am
anthony.green - Monday, November 20, 2017 5:17 AMIt will depends on the quorum options you have defined.You need a witness so disk or fileshare usually and that will determine what would happen if you get failover or shutdown of the services.
Thanks Anthony.
We have a disk witness quorum. But will that make sure the node moves to data center were network is available? What if the Quorum Disk in in Data Center 1 when the connection between Data Center goes off and also the connectivity to that Data Center
November 20, 2017 at 8:04 am
VastSQL - Monday, November 20, 2017 7:54 AManthony.green - Monday, November 20, 2017 5:17 AMIt will depends on the quorum options you have defined.You need a witness so disk or fileshare usually and that will determine what would happen if you get failover or shutdown of the services.
Thanks Anthony.
We have a disk witness quorum. But will that make sure the node moves to data center were network is available? What if the Quorum Disk in in Data Center 1 when the connection between Data Center goes off and also the connectivity to that Data Center
Can that Quorum Disk migrate to Data Center 2? If so you will still achieve quorum. If not you wont have quorum and the cluster will shutdown.
November 20, 2017 at 10:48 am
anthony.green - Monday, November 20, 2017 8:04 AMVastSQL - Monday, November 20, 2017 7:54 AManthony.green - Monday, November 20, 2017 5:17 AMIt will depends on the quorum options you have defined.You need a witness so disk or fileshare usually and that will determine what would happen if you get failover or shutdown of the services.
Thanks Anthony.
We have a disk witness quorum. But will that make sure the node moves to data center were network is available? What if the Quorum Disk in in Data Center 1 when the connection between Data Center goes off and also the connectivity to that Data Center
Can that Quorum Disk migrate to Data Center 2? If so you will still achieve quorum. If not you wont have quorum and the cluster will shutdown.
You mean to say we need to always keep the quorum disk on the passive node? Right now both both A\A and A\P clusters the quorum is in Node 1 which is in Data Center 1.
November 21, 2017 at 1:36 am
VastSQL - Monday, November 20, 2017 10:48 AManthony.green - Monday, November 20, 2017 8:04 AMVastSQL - Monday, November 20, 2017 7:54 AManthony.green - Monday, November 20, 2017 5:17 AMIt will depends on the quorum options you have defined.You need a witness so disk or fileshare usually and that will determine what would happen if you get failover or shutdown of the services.
Thanks Anthony.
We have a disk witness quorum. But will that make sure the node moves to data center were network is available? What if the Quorum Disk in in Data Center 1 when the connection between Data Center goes off and also the connectivity to that Data Center
Can that Quorum Disk migrate to Data Center 2? If so you will still achieve quorum. If not you wont have quorum and the cluster will shutdown.
You mean to say we need to always keep the quorum disk on the passive node? Right now both both A\A and A\P clusters the quorum is in Node 1 which is in Data Center 1.
No, if the quorum disk is in the 2nd data center, and you loose the 2nd data center you have lost quorum again.
HA/DR setups can be tricky, you always need to ensure you have a majority vote.
So
Node1 in DC1, Node2 in DC2 and a disk witness that can failover between DCs, that way if the disk can fail over between DC's you always have 2 of 3 votes which is majority and the cluster roles stay online
Node1 in DC1, Node2 in DC2 and a disk witness which CANNOT failover between DC's, if the disk is in DC1 and DC1 goes offline the cluster goes offline, if the disk is in DC2 and DC2 goes offline the cluster goes offline. 1 of 3 votes is NOT majority and the cluster shuts down
You may be better looking at fileshare witness with a fileshare in another DC or out in the cloud somewhere, that way you are not reliant on a particular DC to keep your clusters alive.
You could also look at forcing a cluster online after loss of quorum but again that could lead to data loss.
November 22, 2017 at 5:07 am
anthony.green - Tuesday, November 21, 2017 1:36 AMVastSQL - Monday, November 20, 2017 10:48 AManthony.green - Monday, November 20, 2017 8:04 AMVastSQL - Monday, November 20, 2017 7:54 AManthony.green - Monday, November 20, 2017 5:17 AMIt will depends on the quorum options you have defined.You need a witness so disk or fileshare usually and that will determine what would happen if you get failover or shutdown of the services.
Thanks Anthony.
We have a disk witness quorum. But will that make sure the node moves to data center were network is available? What if the Quorum Disk in in Data Center 1 when the connection between Data Center goes off and also the connectivity to that Data Center
Can that Quorum Disk migrate to Data Center 2? If so you will still achieve quorum. If not you wont have quorum and the cluster will shutdown.
You mean to say we need to always keep the quorum disk on the passive node? Right now both both A\A and A\P clusters the quorum is in Node 1 which is in Data Center 1.
No, if the quorum disk is in the 2nd data center, and you loose the 2nd data center you have lost quorum again.
HA/DR setups can be tricky, you always need to ensure you have a majority vote.
So
Node1 in DC1, Node2 in DC2 and a disk witness that can failover between DCs, that way if the disk can fail over between DC's you always have 2 of 3 votes which is majority and the cluster roles stay onlineNode1 in DC1, Node2 in DC2 and a disk witness which CANNOT failover between DC's, if the disk is in DC1 and DC1 goes offline the cluster goes offline, if the disk is in DC2 and DC2 goes offline the cluster goes offline. 1 of 3 votes is NOT majority and the cluster shuts down
You may be better looking at fileshare witness with a fileshare in another DC or out in the cloud somewhere, that way you are not reliant on a particular DC to keep your clusters alive.
You could also look at forcing a cluster online after loss of quorum but again that could lead to data loss.
Thanks a lot Anthony.
What about the disk in a cluster storage which can move between the two Data Centers?
We have added a third node for AO and I was planning to change the Quorum configuration to Node Majority but again the problem is third node will be always in Data center 3 .This configuration can also affect cluster if the network goes down.
Can you suggest what should be the quorum configuration when we have 3 Node ? Right now its 3 Node and 1 Disk Quorum Vote (Node and Disk ) all have vote and its even now.
November 22, 2017 at 6:30 am
If the disk can migrate between all the nodes in the cluster and the nodes are all in separate data centers then your OK as you will always achieve quorum.
For a 3 node cluster, it would be better to be node majority only, but again you need to think about node placement
2 nodes in DC1 and 1 node in DC2 is bad. Lose DC1 you lost quorum, even with a disk vote you have lost quorum as either node (1 from 3 vote) or node & disk (2 from 4 votes) is not considered majority.
1 node in DC1, 1 node in DC2, 1 node in DC3 is better as you can sustain a DC loss, lose 2 DC's you've lost quorum, the likely hood of losing 2 DC's at the same time, depends on how you setup the network or how close together the DC's are in the event of a geographical outage.
November 22, 2017 at 9:18 pm
anthony.green - Wednesday, November 22, 2017 6:30 AMIf the disk can migrate between all the nodes in the cluster and the nodes are all in separate data centers then your OK as you will always achieve quorum.For a 3 node cluster, it would be better to be node majority only, but again you need to think about node placement
2 nodes in DC1 and 1 node in DC2 is bad. Lose DC1 you lost quorum, even with a disk vote you have lost quorum as either node (1 from 3 vote) or node & disk (2 from 4 votes) is not considered majority.
1 node in DC1, 1 node in DC2, 1 node in DC3 is better as you can sustain a DC loss, lose 2 DC's you've lost quorum, the likely hood of losing 2 DC's at the same time, depends on how you setup the network or how close together the DC's are in the event of a geographical outage.
Thanks a lot Anthony. Appreciate your help.
We don't have a third data center at the moment. Our servers are VM and are designed in such a way that if server fails it will reboot in same DC in another ESX host if all host fails it will reboot in another DC. My concern is more about losing connectivity between 2 Data Centers and connectivity to any particular DC, say DC1 lost connectivity and if heart beat fails then the cluster will fail right instead of moving all resource to DC 2.
November 23, 2017 at 1:06 am
Unable to comment as that would be up to how your VMware guys have setup the infrastructure, how your network admins setup the network and how you are doing business continuity and disaster recovery solutions for your needs.
If your reliant on vMotion to move all your resources from DC1 to DC2, then you have no concerns over achieving quorum as you should get the exact same machines vMotion'ed from DC1 to DC2, so you wont actually need any servers in DC2 as VMWare will take care of it for you.
You'll be best of going back to all parties involved, getting your RTO/RPO and looking if your current setup can meet those requirements.
November 24, 2017 at 6:46 am
using cross site shared storage for the quorum is a bad idea tbh.
If your cluster spans sites consider a fileshare witness on a remote site (not the primary or secondary site).
Firstly, please confirm the operation system of the windows servers involved in the WSFC?
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
Viewing 11 posts - 1 through 10 (of 10 total)
You must be logged in to reply to this topic. Login to reply