January 7, 2008 at 7:23 am
I have a customer who may want to introduce a SQL cluster to his environment, where one node of the cluster is in one location, and the other node is in a geographically separate location.
My question is: Is there any "distance limitation" with regards to the nodes in a SQL cluster. Is there any recommendations from Microsoft about the placement of these two machines? I have not be able to find any sites that have distances or connection speeds with regards to this kind of setup.
January 8, 2008 at 5:59 am
I'm a relative noob myself, but I have half a dozen clusters. They all have in common that the data drives, log drives, quorum drives, etc. are shared by the different nodes. Thus you would want both nodes to be close to the SAN or drive array.
So, IMHO, having the nodes far apart geographically might be possible (I can't confirm) but the performance would really stink if the active node was the one not physically close to the shared drives/volumes.
Mike Hinds Lead Database Administrator1st Source BankMCP, MCTS
January 8, 2008 at 8:44 am
I have worked with Majority Node Set clusters where the nodes were located in locations over 100kms form each other. This was done using mirrored SAN disks, each node writing to it's local disks and those disks mirrored to the remote site. Only the local disks would be writable untill such time as a failover occured, a script that was hosted in the cluster as a clustered resource, would then make the local disks readonly and the remote disks writable.
The most common vender for this type of solution I've seen is Hitachi, resold by SUN.
January 8, 2008 at 2:14 pm
google "Geographically Dispersed Clusters". Here is MS doc http://download.microsoft.com/download/7/6/f/76f3db2f-6f43-4624-bfde-ff731e3c1f96/GDClusters.doc#_Toc87754923
EMC Geospan/SDRF has remote replication solutions for this. You may look into this.
I remember the distance is 2 digits number, maybe 60KM or some number less than 100 KM. Just could not find the exact document now.
January 8, 2008 at 2:37 pm
holly_westell (1/7/2008)
I have a customer who may want to introduce a SQL cluster to his environment, where one node of the cluster is in one location, and the other node is in a geographically separate location.My question is: Is there any "distance limitation" with regards to the nodes in a SQL cluster. Is there any recommendations from Microsoft about the placement of these two machines? I have not be able to find any sites that have distances or connection speeds with regards to this kind of setup.
Probably a better solution would be to use database mirroring over a WAN.
You may want to look at Tara Kizer's blog over at SQLTeam. I believe she implemented data mirroring over a WAN link with a cluster on each end of the WAN.
http://weblogs.sqlteam.com/tarad/
You may want to read this also:
Database Mirroring Best Practices and Performance Considerations
http://www.microsoft.com/technet/prodtechnol/sql/2005/technologies/dbm_best_pract.mspx
January 8, 2008 at 2:48 pm
Vivien Xing (1/8/2008)
google "Geographically Dispersed Clusters". Here is MS doc http://download.microsoft.com/download/7/6/f/76f3db2f-6f43-4624-bfde-ff731e3c1f96/GDClusters.doc#_Toc87754923EMC Geospan/SDRF has remote replication solutions for this. You may look into this.
I remember the distance is 2 digits number, maybe 60KM or some number less than 100 KM. Just could not find the exact document now.
there's always one,,,,,,,,,,,,,,,,
3x MNS Clusters with nodes in Luxembourg and Mechelen, distance 233Kms+
Please feel free to email me and i'll give you the organisations name and a telephone number you can call and ask them your selves ;o)
lol
January 8, 2008 at 3:13 pm
On training courses and in HA seminars I have always seen the figure of 100 miles quoted as a max for nodes in a MS cluster. The limitation would be getting a quick enough response from the 'heartbeat'
---------------------------------------------------------------------
January 9, 2008 at 4:14 am
The maximum distance between geographically dispersed nodes has gone up over time as technology has improved. This relates both to the software involved in dealing with the heartbeat and the network infrastructure.
My old place is implementing synchronous replication between a large hosted vendor sites in south England and the middle of Belguim (approx 250 miles). The network vendor did tests using drums of optical fibre, sending the signal round and round to so it covered more than the distance between the sites, and it proved workable. It it all goes belly-up when run live you will see it on the evening news.
Original author: https://github.com/SQL-FineBuild/Common/wiki/ 1-click install and best practice configuration of SQL Server 2019, 2017 2016, 2014, 2012, 2008 R2, 2008 and 2005.
When I give food to the poor they call me a saint. When I ask why they are poor they call me a communist - Archbishop Hélder Câmara
January 9, 2008 at 5:09 am
Having experienced Replication over a WAN, I would recommend mirroring over clustering as it should provide more resilience. The WAN I had to operate with constantly failed in bad weather. But I suppose it depends on how reliable the WAN you are working with is.
January 9, 2008 at 10:29 am
Found EMC doc for the distance. For your reference.
Quote from EMC SRDF/CE V2.0 user guide:
“EMC GeoSpan for MSCS provides disaster-tolerant capabilities that enable the cluster servers to be geographically separated by distances of up to 200 KM or where the network connection provides a guaranteed maximum round-trip latency between nodes of no more than 300 milliseconds."
October 7, 2008 at 9:21 am
Argggg...
I 'think' I need to set up a cluster with nodes in different cities for dr.
I have a lot of experience with clustering in the same location but none over long distances.
If I have a 3 node cluster with 2 nodes in location A and one in location B connected to a SAN, and one node in location A goes away other node in location A picks up without issue. What happens if location A goes away???
How do I get the data to location B if the SAN was located at A?
Is there a way to 'cluster' the SAN between locations?
Thanks in advance
October 7, 2008 at 9:32 am
Michael.varriale (10/7/2008)
Argggg...I 'think' I need to set up a cluster with nodes in different cities for dr.
I have a lot of experience with clustering in the same location but none over long distances.
If I have a 3 node cluster with 2 nodes in location A and one in location B connected to a SAN, and one node in location A goes away other node in location A picks up without issue. What happens if location A goes away???
How do I get the data to location B if the SAN was located at A?
Is there a way to 'cluster' the SAN between locations?
Thanks in advance
if you lose the whole 'location' where the SAN the cluster is attached to is located, you lose your cluster, hence clustering is more high availability than a full blown DR soulution. You need to look at SRDF possibly to replicate the SAN, or even log shipping or mirroring on top of the cluster.
---------------------------------------------------------------------
December 4, 2008 at 6:56 am
We are also looking at this type of scenario, is there any information on changing a standard two node cluster into a three node cluster with the quorum device also changing from a stanadrd qurorum device to a majority node quorum.
February 10, 2011 at 7:38 am
Hello,
I have been trying to find information of building sql clusters across a WAN with 1Gb of throughput between two datacenters. The cluster is currently built against a san sitting at one site and we wish to either load balance across two sites or will settle with a failover against the back end data connection. There is a san at the second site.
From the article you have written you mimic our configuration quite closely and was wondering if you had an architecture docs i can copy?
February 10, 2011 at 7:40 am
Iain-377297 (1/8/2008)
I have worked with Majority Node Set clusters where the nodes were located in locations over 100kms form each other. This was done using mirrored SAN disks, each node writing to it's local disks and those disks mirrored to the remote site. Only the local disks would be writable untill such time as a failover occured, a script that was hosted in the cluster as a clustered resource, would then make the local disks readonly and the remote disks writable.The most common vender for this type of solution I've seen is Hitachi, resold by SUN.
Can someone direct me to some architectural documentation similar to what iain has written as i have an environment that is very similar and am looking to conduct the same setup?
Viewing 15 posts - 1 through 15 (of 16 total)
You must be logged in to reply to this topic. Login to reply