Understanding Server 2012r2 Clustering
-
@dafyre said:
@scottalanmiller -- Just so I understand... In most cases, Application Level Clustering > Windows Failover Clustering ?
Probably in all cases but there must be one where this isn't true. But conceptually, application level clustering is the only way to get true, completely reliable failover (when done right.) Anything else is an attempt to make up for lacking application clustering. Windows Failover, VMware failover, etc. are all "making do", not ideal.
-
Exchange is one time you should never use a SAN. Nor can you use Vmotion with Exchange. If you are running Exchange on site most of the time you might as well look at separate physical boxes but, then that comes down too why are you looking at exchange onsite vs hosted?
-
@thecreativeone91 said:
Exchange is one time you should never use a SAN. Nor can you use Vmotion with Exchange. If you are running Exchange on site most of the time you might as well look at separate physical boxes but, then that comes down too why are you looking at exchange onsite vs hosted?
Other times include MS SQL Server (or pretty much any database), Active Directory, etc. Anything that has an open data connection.
-
The times that SAN can be used for a reliably consistent failover are actually pretty rare and almost always cases where there was an easy way to have done it without a SAN.
-
@scottalanmiller I would argue that about MSSQL and MySQL. We ran those on the Same box (as part of the same cluster) for a number of years. The only minor issue that would happen is that the SIS that the Campus used would throw an error message and wouldn't automatically reconnect. The error message I can understand. But not automatically reconnecting? That is an application issue and not a problem with Failover.
Our MySQL applications never had this problem.
We were probably just lucky, but we never lost any data in MSSQL Server due to a failover event.
-
@dafyre said:
@scottalanmiller I would argue that about MSSQL and MySQL. We ran those on the Same box (as part of the same cluster) for a number of years. The only minor issue that would happen is that the SIS that the Campus used would throw an error message and wouldn't automatically reconnect. The error message I can understand. But not automatically reconnecting? That is an application issue and not a problem with Failover.
Our MySQL applications never had this problem.
We were probably just lucky, but we never lost any data in MSSQL Server due to a failover event.
MS SQL Server, MySQL, MariaDB, Oracle DB, PostgreSQL, DB2, Sybase... you name it. They can't survive having their storage ripped out from under them. SAN = violent storage ripping.
-
Server 2012 Has safeguards in place. It's fine to run DCs on a SAN and use vMotion with Server 2012 or newer. The VM Generation ID is there for this reason. Even Cloning of DCs is now supported and if done properly will not cause USN issues.
-
@thecreativeone91 said:
Exchange is one time you should never use a SAN. Nor can you use Vmotion with Exchange. If you are running Exchange on site most of the time you might as well look at separate physical boxes but, then that comes down too why are you looking at exchange onsite vs hosted?
I never knew this. What's the problem with Vmotion and/or failover with Exchange? What's the problem with SQL Server?
Not that I have a SAN or HA, I'm just interested. I was interested at the time I considered a SAN (a few years ago) by the fact that my reseller recommend against DAGs on the grounds of cost (additional licencing) and complexity, but recommended in favour of a SAN (which also has additional costs and complexity). At the time I couldn't understand their reasoning. I was always more inclined towards application level clustering - it just seemed to make more sense.
Exchange and SQL Server are both designed to recover from a crash. At worst, shouldn't failover be at least as good as a crash consistent recovery?
-
@Carnival-Boy said:
I never knew this. What's the problem with Vmotion and/or failover with Exchange? What's the
problem with SQL Server?The problem is the state of the open files and open connections.
Exchange and SQL Server are both designed to recover from a crash. At worst, shouldn't failover be at least as good as a crash consistent recovery?
Yes, but vMotion/Live Migration is something you do that is not a crash scenario. It is a day to day business process.
Paid versions of Veeam have options to be aware of items in the guest and can easily move these types of services as ling as you know that any data from an open connection may get lost. Conveniently, most connections are simultaneous calls to the database and closed again. Only poorly designed applications anymore hold connections open.
-
Of course they recommended a SAN, they make a ton of money off the sale, conversely they barely make anything selling you multiple copies of Exchange for the DAGs.
I'm wondering when or if we'll see MSPs, move away from the bad recommendations SANs anytime soon?
I had a friend who started at a small school district, they needed a new server. Their vendor/local computer shop sold them a one box VM host with a SAN. When he told me that I freaked out on him... told him why this was a horrible solution - he didn't seem to care. "It's to late" he said, "It's already done and installed and working."
Even now when I talk to him he doesn't seem to understand why this is bad.
-
@Dashrender said:
Of course they recommended a SAN, they make a ton of money off the sale, conversely they barely make anything selling you multiple copies of Exchange for the DAGs.
soon?^^^ Cannot be overstated. Sales people are paid, by you, you sell the things that make them money. Not that they won't sell you other things, but when you are talking about something that is orders of magnitude more money in profits for them, you can't expect them to voluntarily do engineering work that you are not paying them for instead of selling you the high margin item that you are paying them for.
-
@Dashrender said:
I'm wondering when or if we'll see MSPs, move away from the bad recommendations SANs anytime soon?
This is never going to be something that VARs (not MSPs) will change voluntarily. This is something that has always been driven by the customers. Only when the customers start paying for advice and not trying to get free advice lumped in with sales will this change. Most VARs are exclusively compensated by selling high margin solutions, they are not paid to provide good advice. If we ask for free advice, someone has to pay for that advice. So we, as the customers, determine by trying to get free advice which solutions the VAR will be paid for and which they will not.
Customers choose how their vendors are engaged. Vendors can choose not to work with customers in a certain manner, but the customer will always find a sales person willing to make a sale. As long as customers work that way, companies will exist to support that desire.
-
@Carnival-Boy said:
Not that I have a SAN or HA, I'm just interested. I was interested at the time I considered a SAN (a few years ago) by the fact that my reseller recommend against DAGs on the grounds of cost (additional licencing) and complexity, but recommended in favour of a SAN (which also has additional costs and complexity).
Remember the long conversation that we had about how you trust your sales people and don't feel that the financial compensation for selling certain solutions was influencing them and that they acted altruistically? This is the text book example that we use for what a vendor taking advantage of you looking for free advice looks for. The article that I wrote was because of the prevalence of this exact case (the Inverted Pyramid of Doom SAN sales tactic.)
This is what what I was warning you about looks like. This is a case where it is insanely obvious that what the vendor was trying to sell was completely illogical, and luckily you caught that, but by asking a salesperson for architectural advice you were both asking them something that they were not paid to understand and something outside of their likely skill set and also to give you advice that if they tell you what is good for you they don't get paid and if they tell you what you pay them to tell you, the advice is reckless.
This gets mentioned over on SW a few times a week, about how vendors specifically use the IPOD design because it is really easy to trick management into feeling that it is reliable because the words "redundant" and "SAN" appear while the vendor makes big profits, keeps the purchase price down and passes all risk on to the customer. It's the most common example of why getting advice from the vendor is a dangerous thing to do because their interests and your interests are not aligned and they have absolutely no obligation to giving good advice that is in your interest because no such paid professional advice relationship exists or, even if one does, as a sales person it is absolutely understood that they are compensated for making profitable sales regardless of any other additional relationship.
-
@Carnival-Boy said:
Exchange and SQL Server are both designed to recover from a crash. At worst, shouldn't failover be at least as good as a crash consistent recovery?
Most things are designed to recover from a crash. That doesn't mean that you won't lose some data or that it always works. Exchange and SQL Server are quite robust, but failover that acts like you've had a disaster isn't a full failover.
For most companies, finding out that their failover is "only as good as a crash" is quite upsetting.
-
@Dashrender said:
When he told me that I freaked out on him... told him why this was a horrible solution - he didn't seem to care. "It's to late" he said, "It's already done and installed and working."
This is one of those excuses that really bad IT pros seem to use a lot. "It is working." They redefine what "working" means as it suites them. It's like driving around without a seatbelt on. Sure, as long as you are driving, it is "working." But the seatbelt is there for when you hit a tree. When you go through the windshield and lie bleeding out on the ground, will you still say "it is working?"
The point of good design, in this particular base, is to get better reliability and speed at a lower price. If the IT pro defines what he has done as "working" it means that being cost effective and protecting the business financially aren't part of his job. What exactly IS his job then? It sounds like he feels his job is to funnel company money to a vendor and not to provide good IT advice or service. In fact, by using the vendor to do his job for him, he likely didn't do his job as the IT pro at all!
That things are "too late" is one thing. But claiming that it doesn't matter, fixing a discovered problem isn't important or that it is working highlights what he feels his job is or how little he cares about the company he is working for. He's not trying to do the right thing, he is hoping not to get found out that he scammed the business and tried to get away without providing any technical expertise and letting a sales person hoodwink him while he thought he was getting away without having to know his job.
-
It's amazing, given how obviously illogical, how many discussions and how much documentation there is out there about the foolishness of the single SAN Inverted Pyramid of Doom design that any VAR would take the risk of recommending it knowing that tons of IT pros will know instantly that they are being scammed and that they should drop the vendor like a hot rock. But, the reality is, if a VAR is being asked for advice, they know that the customer is trying to get away without doing the research and isn't sure what to do and simply by being asked for the advice know that they can get away with some pretty crazy stuff. So the group of people that will call out the vendors for doing this stuff are the ones who naturally don't engage the vendors in this way making it very safe for the vendors to try so pretty crazy sales tactics and approaches.
-
The flip side to this is that asking the VAR is considered the research to most of them. And really, until I joined SW several years ago I was doing exactly the same thing.
-
@Dashrender said:
The flip side to this is that asking the VAR is considered the research to most of them.
In the SMB there seems to be this really broad acceptance of using sales people as IT advisers. It is very widespread. I never saw it happen until being in SW. Literally never. I had no idea that it was done at all so came as a pretty big shock.
It's not IT specific in any way, using sales people for advice is just problematic.
-
Check out SAN topics all the advice to get them are from Sellers, and the HP one especially is full of BS. Yet they still go with the sales peoples advice. The HP one says the SANs of two aren't the same as your dad's SAN, they are much more reliable or something like that.
-
@thecreativeone91 said:
Check out SAN topics all the advice to get them are from Sellers, and the HP one especially is full of BS. Yet they still go with the sales peoples advice. The HP one says the SANs of two aren't the same as your dad's SAN, they are much more reliable or something like that.
I blame management. Instead of looking for their IT staff to provide IT services, they make them just people who buy things from others who provide IT. It's the IT oursource model. Except they refuse to pay for professional IT services. So instead they go to vendors and get advice from the sales people for free and say "this is just the cost of what we need." It's an illusion that is created by bad management. They don't get the skills that they need internally and refuse to let people buy advice, which is really 95% of what IT actually is. Anyone can do the physical parts of IT.
So once management does this, they create a situation where the internal staff isn't sure what to do and isn't able to hire the IT people that they need (externally) and are stuck getting free advice which actually costs them a fortune in risk, bad advice and over purchasing.
The internal staff has no reason to give advice, even if they know better, because it just adds risks to themselves. Why take on risk to save a company money that obviously doesn't respect IT and doesn't care about making money? Why should IT take on that risk? So they don't. They let a vendor oversell and then have the benefit of having someone to "blame" when the company loses money or goes under. How can it be IT's fault when their only job is to sign off on whatever a sales person sold to them?
It's management making IT a "purchaser" rather than a "professional" that creates this situation. If management expected IT to research, give advice and protect the business this really could not happen. But when IT's job is just to talk to the sales people on behalf of management, what else could realistically happen? Once again, management makes their internal staff not be aligned with the company's goals. Why would IT work against its own interests when management are the ones who created the conflict of interests?