• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

vSphere Design - VMware Infratructure DB Best Practice

V

Vader

Supreme [H]ardness
2FA
Joined
Dec 22, 2002
Messages
5,137
As I continue on my training for the Design exam, i'm seeing a common theme of DB separation from the actual Infrastructure application, ie vCenter, VUM, View, and so on and so on.

My questions are pertaining the common protection schemes for the DB's of the VMware INF application tier. Assuming I am picking a DB product, such as SQL for my Databases for these apps,and also assume I have at least 3 or 4 INF Applications that require DB's, would this be considered the best practice design:

1. Centralized DB MS Failover Cluster, RDM's, and VM to host affinity. Build instances off of the same core DB Failover cluster for each application vCenter, VUM, View, etc.

2. INF Applications themselves would sit in the same cluster with normal HA/DRS rules in place.


A lot of the design information i'm gathering is also a bit confusing. As you know you certainly don't want to price your solution out of the ballpark where your customer gets sticker shock. I would think that the above would fly for an enterprise type solution, but how would you handle a smaller business who wants VDI for example? I don't mean like small office, 30 people..etc..but a business that has say 200 people, 150 of those would be VDI. Would you still consider the MGMT cluster and centralized DB type solution? I'm guessing yes looking at the POD and BLOCK design guideline.
 
Last edited:
What about integration into an existing SQL Cluster? I would venture that organizations that would consider a SQL cluster for their VMware INF probably already has a SQL cluster setup for their other production databases and might want to leverage that instead of building a new cluster. Sorry if that's slightly off-topic but I thought it might be worth throwing in there as well.
 
Standard database availability practices are in place. There's nothing wrong with having a MSSQL failover cluster in the same vSphere cluster as vCenter, VUM, vCM, vCD, etc. Common sense obviously says DRS anti-affinity rules for the two virtual SQL nodes. There are published white papers for SQL and Oracle availability.

Make sure you are aware of vCenter heartbeat as well.

I don't remember getting anything obscure when it came to DB design measures on the DCD.
 
What about integration into an existing SQL Cluster? I would venture that organizations that would consider a SQL cluster for their VMware INF probably already has a SQL cluster setup for their other production databases and might want to leverage that instead of building a new cluster. Sorry if that's slightly off-topic but I thought it might be worth throwing in there as well.
Click to expand...

It's certainly not off topic, and I think yeah, that makes sense. If there is already a core DB infrastrucutre where the product is supported for the INF Applications, then absolutely, didn't really think of that.

Make sure you are aware of vCenter heartbeat as well.

I don't remember getting anything obscure when it came to DB design measures on the DCD.
Click to expand...

Let's discuss that. vCenter Heartbeat provides a redundant vCenter but I hear it's at a pretty steep cost, i've seen $12k? Not sure how accurate that is. This would also require another DB instance i'm assuming and I guess you also have to factor in additional network capacity for the replication to occur etc?

This is not just about the DCD for me, it also has to do with a planned career change from an operational IT role to a pre-sales/implemenation role.:)
 
Vader said:
Let's discuss that. vCenter Heartbeat provides a redundant vCenter but I hear it's at a pretty steep cost, i've seen $12k? Not sure how accurate that is. This would also require another DB instance i'm assuming and I guess you also have to factor in additional network capacity for the replication to occur etc?

This is not just about the DCD for me, it also has to do with a planned career change from an operational IT role to a pre-sales/implemenation role.:)
Click to expand...

You heard correctly. I'm not going to say it is worthless but it still has the same dependencies with some extra moving parts (at a pricetag, I've heard it was 10k.. but I'm sure it could be 12 with support bundled in). I was just saying to be more aware of the product as an additional availability solution

Have you had the chance to read Scott Lowe's vSphere Design book? It was written in the 4.1 days but everything in there is still relevant to design/implementation considerations.
 
Not a fan at all of MS Clustering and see no need for it in VMware environments. Let HA do the work without the complexities and trouble.

Very few people buy vCenter Heartbeat. It's expensive and again....doesn't add a whole lot for the money. Let HA do the work for you.

I like management clusters but you only see them in big environments, much larger than the examples OP gives. And really...they aren't needed in a "traditional" server consolidation cluster. Add in VDI or vCloud Director and that's different...but almost everyone just has vCenter/DB/whatever in the same cluster as server workloads since they would be protected the same. We split out management from VDI for licensing reasons often (vSphere for VDI says no servers). For vCloud it's best to split it out for management function and not to mix management and resources.
 
Now that I have my coffee... (Sitting at Starbucks before heading over to VMware HQ)... If this is an exercise for say a pre-sales thing.. You can't put down requirements without putting down budget. Designs are easy in absence of constraints. The reason you don't see management clusters as often as we'd like is budget.

Let's say it was 50 servers and 150 VDI. 150 VDI is often 3 servers in their own cluster. Servers...well, 50 is only 2 servers usually..add a third for N+1. Want a management cluster? Minimum of 2 servers. So you've added two more servers PLUS four more vSphere Enterprise or Enterprise Plus licenses and SnS on those against maybe one more server in the server cluster.

Every time you add a cluster you lose efficiency due to N+1 or N+2 capability plus licensing plus support...etc.

It's what I call the "new pre-sales guy trap". They design toward best practices..but best practices aren't always best practices.
 
Not a fan at all of MS Clustering and see no need for it in VMware environments. Let HA do the work without the complexities and trouble.

Very few people buy vCenter Heartbeat. It's expensive and again....doesn't add a whole lot for the money. Let HA do the work for you.

I like management clusters but you only see them in big environments, much larger than the examples OP gives. And really...they aren't needed in a "traditional" server consolidation cluster. Add in VDI or vCloud Director and that's different...but almost everyone just has vCenter/DB/whatever in the same cluster as server workloads since they would be protected the same. We split out management from VDI for licensing reasons often (vSphere for VDI says no servers). For vCloud it's best to split it out for management function and not to mix management and resources.
Click to expand...

Yeah..I was thinking it probably could get pretty pricey going this route.

Let me ask this question, then, you still setup a core DB infrastructure, correct, except it's just another VM and uses VMFS volumes for the DB's? I would assume a single Server running Oracle or SQL and running multiple instances for the different products, ie vCenter, VUM..etc?

Have you had the chance to read Scott Lowe's vSphere Design book? It was written in the 4.1 days but everything in there is still relevant to design/implementation considerations.
Click to expand...

I'm about 1/2 way through now. I found some other material from a VMWare presentation that's rather interesting from here:

http://virtual-red-dot.blogspot.com/

He also has some interesting views on design.

I definately understand that everything is pending customer requirements and budget..I guess I just want to make sure that i'm going dwon the right path and would like to have some sort of standards or baseline that I can build from.
 
Last edited:
Yep...that's it. You could do multiple SQL VMs if you'd like. That all comes down to licensing. If they have a lot of SQL it might pay off to license all the CPUs/cores in a cluster and just run a bunch...or it might not. That's a pricing exercise.
 
Yep...that's it. You could do multiple SQL VMs if you'd like. That all comes down to licensing. If they have a lot of SQL it might pay off to license all the CPUs/cores in a cluster and just run a bunch...or it might not. That's a pricing exercise.
Click to expand...

Ok..good, then i'm on the right track for that piece. I'm sure there will be many more posts coming...lol.

It looks like I have a solid opportunity at this VAR I posted about earlier that was a strong Cisco/NetApp/VMware partner. I had a foray of interviews yeseterday that went well. The specific needs look to be centered around Flexpod, which i'm very excited about, and VDI using View.

If I don't get some hands on for UCS soon, i'm going to buy the Trainsignal material.
 
If you don't let me know. I can do a Gotomeeting for an hour or two one day and give you some good training..enough to speak very intelligently about it.
 
If you don't let me know. I can do a Gotomeeting for an hour or two one day and give you some good training..enough to speak very intelligently about it.
Click to expand...

That would be awesome...i'll let you know.
 
NetJunkie said:
Not a fan at all of MS Clustering and see no need for it in VMware environments. Let HA do the work without the complexities and trouble.
Click to expand...

Not saying that I'm a fan either, but I'm talking about running the vCenter instance on a sql cluster that is also the home of many other databases that need a highly available home as we all know, HA != highly available.

I agree with everything else though.

Vader, best of luck with the possible career move... definitely sounds like good stuff!
 
How is MS Clustering more available than HA? MS Clustering isn't active/active.
 
NetJunkie said:
How is MS Clustering more available than HA? MS Clustering isn't active/active.
Click to expand...

MS Clustering is at least application-aware. HA can be argued to be OS aware so I'll leave that point out.

Just so we're on the same page here, I wouldn't go out of my way to make a SQL cluster just to host vCenter. But there are many cases where a customer requires the availability of a SQL cluster, in that event I would suggest vCenter to go there.

We all know HA isn't instant (FDM or legato) so it is not a replacement by any means. I would never go into a customer site and say "oh, you're using a MSCS SQL cluster? don't worry, we'll tear that down because you have HA now" I would have been fired years ago. Most of the time for me the main feature of the MSCS clustering was just to keep up with all of the damn windows and sql patches without outages.

I like phasing out MS stuff just as much as the next guy, but MSCS still does have a place.
 
Really? I tell customers that all the time...the vast majority agree with me and are much happier. I have *1* customer currently deploying MORE Microsoft clusters. It's complexity they don't need. The ONLY reason to do it is to patch...that's it. It takes almost as long to failover an app to another node as it does to restart a VM via HA. Given what you need to do MS Clustering and what it does to your routine vSphere maintenance (vMotion issues) it's not worth it for the vast majority.

I have been asked by a number of customer IT support teams to go argue for them to LoB app owners against MS Clustering and explain that it's just not worth it anymore.
 
Also, sitting here at Partner Technical Advisory Board at VMware...MS Clustering came up during a vCloud Architecture bootcamp and everyone in the room said the same thing..customers are fleeing it very quickly when they virtualize.
 
NetJunkie said:
Really? I tell customers that all the time...the vast majority agree with me and are much happier. I have *1* customer currently deploying MORE Microsoft clusters. It's complexity they don't need. The ONLY reason to do it is to patch...that's it. It takes almost as long to failover an app to another node as it does to restart a VM via HA. Given what you need to do MS Clustering and what it does to your routine vSphere maintenance (vMotion issues) it's not worth it for the vast majority.

I have been asked by a number of customer IT support teams to go argue for them to LoB app owners against MS Clustering and explain that it's just not worth it anymore.
Click to expand...

Don't get me wrong, I do agree with your points. But I can't go as far to say that HA is the end-all replacement in terms of availability for SQL for all customer needs, especially when it is powering BCA's. Plus, when you get the DBA's in the conversation where there is already enough pushback for virtualization that there is a level of comfort being able to maintain a traditional SQL cluster as they know it.

I believe we may have went off on a tangent :) Good stuff though!
 
Last edited:
I can't tell you how great conversations like these are. I believe that are a lot more lurkers here that read these forums that don't necessarily post where they get a lot of useful information.

Also..when you do a google search this link definately comes up so it's out there for a lot more to see.

It's great for people in my position, I don't have coworkers to bounce ideas off of that deal with virtualization...etc, I rely on networking and some of the blogs and forums like these.

Back on topic, in terms of HA for the SQL VMware INF you would want to enable Application monitoring for the SQL DB i'm assuming?
 
Vader said:
I can't tell you how great conversations like these are. I believe that are a lot more lurkers here that read these forums that don't necessarily post where they get a lot of useful information.

Also..when you do a google search this link definately comes up so it's out there for a lot more to see.

It's great for people in my position, I don't have coworkers to bounce ideas off of that deal with virtualization...etc, I rely on networking and some of the blogs and forums like these.

Back on topic, in terms of HA for the SQL VMware INF you would want to enable Application monitoring for the SQL DB i'm assuming?
Click to expand...

I actually found this place via a IPv6 related search that ultimately led me to the fix I needed.

The VMware communities are great but since becoming an employee I got that little icon under my name and I don't care for the implications that come with that, like the perception of being all-knowing (which is far from the truth). So I think I'll stick around here, already some sharp individuals to converse and learn from.
 
You in Palo Alto? We could just argue in person. :) Here for the rest of the evening at the EBC.
 
NetJunkie said:
You in Palo Alto? We could just argue in person. :) Here for the rest of the evening at the EBC.
Click to expand...

Rarely ever, I'm a road warrior.

Enjoy your stay at the main campus, definitely a cool facility.

Annnnnnd just so we get off on the right foot, I pretty much agree with all of your points previously. Just had to throw in my experiences/encounters for an added perspective. I actually just saw your blog link, have been there plenty of times. Keep up the solid work. I think I found it through the Varrow website (I'm a Raleigh resident and was checking out the company that had the only VCDX in the South East)
 
We were on the right foot. :) I like to see different perspectives. Funny how things like this can vary by vertical, geo, etc.
 
NetJunkie said:
We were on the right foot. :) I like to see different perspectives. Funny how things like this can vary by vertical, geo, etc.
Click to expand...

Absolutely. I operate as a lone wolf so mediums like this to converse with colleagues is critical to me.

I attempted to get the clearance to go to the Charlotte VMUG for this same reason, but no dice. Excellent list of speakers lined up though.
 
Doh. I wish you had. I'm speaking on vDS...well, that's the plan. I kind of need to prep that. :)
 
subbed.. I certainly appreciate all of these brilliant minds coming together here :)

Sign me up for the NetJunkie webinar too, if I can't hear you speak at a VMUG this is probably the next best thing!
 
shnelson said:
subbed.. I certainly appreciate all of these brilliant minds coming together here :)

Sign me up for the NetJunkie webinar too, if I can't hear you speak at a VMUG this is probably the next best thing!
Click to expand...

Oh, webinar? I'm in. Is it going to be a dry run of your vDS presentation?
 
itrsteve said:
Oh, webinar? I'm in. Is it going to be a dry run of your vDS presentation?
Click to expand...

Sorry, didn't mean to mislead with that last post... I saw NetJunkie offer a gotomeeting session with Vader & was trying to piggyback. Not sure what the intent or content is going to be, I just know I enjoy absorbing the knowledge from within his blog and could surely benefit from a listen/watch session no matter the subject.
 
We could do that sometime. Little group session on vDS and/or UCS and other stuff.
 
Would need to start after EMC World and I'm done with this Nexus 1000v TrainSignal course.
 
I just started supporting a pretty large cluster using the vDS and we are also hoping to sell them UCS soon, I would really appreciate it if could attend as well. Not nearly as experienced as you guys but I learn pretty fast :)
 
We have multiple SQL clusters in our environment and yes it is a pain in the ass for maintenance but how are you providing OS/App level redundancy with only HA? What if the OS crashes?
 
Zlash said:
We have multiple SQL clusters in our environment and yes it is a pain in the ass for maintenance but how are you providing OS/App level redundancy with only HA? What if the OS crashes?
Click to expand...

Heartbeats from VMware Tools stop being received and the VM is restarted on another host.
 
Zlash said:
Yea but that doesn't help if the OS is hosed and can't boot or whatever though.
Click to expand...

What backups are for. That's really not the use case for MS Clustering...though it does protect against it...that's not the reason 90%+ of people deploy it. They do it for app resiliency. What hoses an OS? If anything it's a new patch. In a virtual world I snapshot all my VMs before I patch and test. If it hoses I roll it back.
 
NetJunkie said:
What backups are for. That's really not the use case for MS Clustering...though it does protect against it...that's not the reason 90%+ of people deploy it. They do it for app resiliency. What hoses an OS? If anything it's a new patch. In a virtual world I snapshot all my VMs before I patch and test. If it hoses I roll it back.
Click to expand...

All of that requires downtime though which isn't acceptable to some customers if it can be avoided by having a cluster. Anyway just making sure I wasn't missing something.
 
Agreed. It's all about requirements that feed in to your design. Most people THINK they need 100% uptime but don't. The majority of people rolling out MS Clusters for things don't really need it...they do it because they think they do and end up causing themselves more trouble than it's worth. That's where I come in.
 
You must log in or register to reply here.
Back
Top