New Project - Thoughts? (CentOS, HAProxy, Load Balance)...
-
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Also, for things like Windows Updates. I can take VM 1 out of the load balancer at say midnight, then update the VM, then bring back in... all without any real issues to customers.
Currently, we would affect customers with such things...
Also does not need load balancing
-
Keep in mind that you add an entire extra layer of issues. Sure you make your IIS able to fail over, but you need your load balancer to be able to fail over, too. And you need to be able to take the load balancer offline for updates and so forth.
-
@JaredBusch said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@JaredBusch said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
I don't edit while driving and honestly I missed the word that time
No problem
darn typos!Watch for weird sentence structure from Jared, if you see a weird sentence it almost always means it was a speech to text translator getting words wrong.
Or as in this case if I'm turning around and insulting you for no apparent reason it was probably an accident unless your name is @DustinB3403 and you're talking about XS
Wait, what happened, how am I dragged into this?
-
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Why do you want to load balance multiple IIS instances? Are your individual web servers so big that you can't grow them any longer?
The end goal is to 4 x IIS VMs with each being on a different host. Each VM runs the same site (so 4 running copies on different hardware). With load balancer, especially if they can work as a redundant pair too on separate hosts, we'd have to lose all four IIS VMs, or both Internet lines, or both load balancers, or all four hosts to be unavailable to clients.
Or power/flood/Cloudflare issues/what not.
But in terms of what we have available to us, its within reach to at least have better capability with what we can control. I just need to understand what tools to use now, be in HAProxy, NGINX, NetScaler... etc
But why? What's the benefit to this?
Currently, if IIS website A is running on a VM on Host1, and that host dies, we're down. We have to turn the replica VM on which is on Host2. That's not automatic. By having that website on two VM's, one on Host 1 and one on Host 2, with a load balancer, the lb would stop directing traffic to the failed one and push everything to the live one. The service is better for customers...
Also, for things like Windows Updates. I can take VM 1 out of the load balancer at say midnight, then update the VM, then bring back in... all without any real issues to customers.
Currently, we would affect customers with such things...
This is built into IIS. I'm trying to remember what Microsoft calls it, but you can do it with DFS.
-
@JaredBusch said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@JaredBusch said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
If you're going to constrain yourself to on premise then as Scott suggested Nginx is the probably the best thing to do
We've looked at moving off site a few times. Its way more expensive off site. Since we have our servers, redundant firewalls, switches already, adding a free way to load balance (HAProxy/Nginx/Whatever) adds no real cost to us. We have the hardware and space already.
Moving the kit off site could be done, but why pay to host somewhere else. We have a great DR system, so if we lost Head Office it wouldn't matter. This is for day to day running of the services.
If you already have most of the redundant stuff that you would gain by going to a colocation or a host then yeah you're an exception your companies already have the cost sunk into that so not too bad.
At this point the most cost efficient solution likely will be at implement some kind of low-cost load balancing on site. And look at my greeting offsite again the next time you have to deal with hardware or something else changes the cost equation.
We will look again at the next refresh. I could perhaps see colo as an option with our own hardware being in a datacentre some where like our DR systems... but hosting our VMs on another infrastructure is far more costly and unless prices drop, a lot I don't see that happening.
-
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Currently, if IIS website A is running on a VM on Host1, and that host dies, we're down. We have to turn the replica VM on which is on Host2. That's not automatic. By having that website on two VM's, one on Host 1 and one on Host 2, with a load balancer, the lb would stop directing traffic to the failed one and push everything to the live one. The service is better for customers...
I think you are confusing failover with load balancing. I totally see the benefit to having failover, I don't see any to load balancing. That's why I keep prying. I think that you are looking for the wrong solution. Not that the products aren't the same, but your needs and goals are different.
Load balancing is for when you are too big for one system to host. Until you are that big, load balancing is a negative.
Part of load balancing is to stop directing traffic to down instances right? If
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Also, for things like Windows Updates. I can take VM 1 out of the load balancer at say midnight, then update the VM, then bring back in... all without any real issues to customers.
Currently, we would affect customers with such things...
Also does not need load balancing
Wouldn't load balancing give us what we want here? Load balancer will direct traffic away from a down node. Yes, its not a 'failover', you're right... but customers are not affected...
-
@coliver said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Why do you want to load balance multiple IIS instances? Are your individual web servers so big that you can't grow them any longer?
The end goal is to 4 x IIS VMs with each being on a different host. Each VM runs the same site (so 4 running copies on different hardware). With load balancer, especially if they can work as a redundant pair too on separate hosts, we'd have to lose all four IIS VMs, or both Internet lines, or both load balancers, or all four hosts to be unavailable to clients.
Or power/flood/Cloudflare issues/what not.
But in terms of what we have available to us, its within reach to at least have better capability with what we can control. I just need to understand what tools to use now, be in HAProxy, NGINX, NetScaler... etc
But why? What's the benefit to this?
Currently, if IIS website A is running on a VM on Host1, and that host dies, we're down. We have to turn the replica VM on which is on Host2. That's not automatic. By having that website on two VM's, one on Host 1 and one on Host 2, with a load balancer, the lb would stop directing traffic to the failed one and push everything to the live one. The service is better for customers...
Also, for things like Windows Updates. I can take VM 1 out of the load balancer at say midnight, then update the VM, then bring back in... all without any real issues to customers.
Currently, we would affect customers with such things...
This is built into IIS. I'm trying to remember what Microsoft calls it, but you can do it with DFS.
If you have any details, i'd appreciate seeing them...
-
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Currently, if IIS website A is running on a VM on Host1, and that host dies, we're down. We have to turn the replica VM on which is on Host2. That's not automatic. By having that website on two VM's, one on Host 1 and one on Host 2, with a load balancer, the lb would stop directing traffic to the failed one and push everything to the live one. The service is better for customers...
I think you are confusing failover with load balancing. I totally see the benefit to having failover, I don't see any to load balancing. That's why I keep prying. I think that you are looking for the wrong solution. Not that the products aren't the same, but your needs and goals are different.
Load balancing is for when you are too big for one system to host. Until you are that big, load balancing is a negative.
Part of load balancing is to stop directing traffic to down instances right? If
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Also, for things like Windows Updates. I can take VM 1 out of the load balancer at say midnight, then update the VM, then bring back in... all without any real issues to customers.
Currently, we would affect customers with such things...
Also does not need load balancing
Wouldn't load balancing give us what we want here? Load balancer will direct traffic away from a down node. Yes, its not a 'failover', you're right... but customers are not affected...
The goal of fail-over is to allow work that would normally be done by one server to be done by another server should the regular one fail.
Load balancing lets you spread load over multiple servers. You would want to do this if you were maxing out your CPU or disk IO or network capacity on a particular server.
-
@Romo said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Currently, if IIS website A is running on a VM on Host1, and that host dies, we're down. We have to turn the replica VM on which is on Host2. That's not automatic. By having that website on two VM's, one on Host 1 and one on Host 2, with a load balancer, the lb would stop directing traffic to the failed one and push everything to the live one. The service is better for customers...
I think you are confusing failover with load balancing. I totally see the benefit to having failover, I don't see any to load balancing. That's why I keep prying. I think that you are looking for the wrong solution. Not that the products aren't the same, but your needs and goals are different.
Load balancing is for when you are too big for one system to host. Until you are that big, load balancing is a negative.
Part of load balancing is to stop directing traffic to down instances right? If
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Also, for things like Windows Updates. I can take VM 1 out of the load balancer at say midnight, then update the VM, then bring back in... all without any real issues to customers.
Currently, we would affect customers with such things...
Also does not need load balancing
Wouldn't load balancing give us what we want here? Load balancer will direct traffic away from a down node. Yes, its not a 'failover', you're right... but customers are not affected...
The goal of fail-over is to allow work that would normally be done by one server to be done by another server should the regular one fail.
Load balancing lets you spread load over multiple servers. You would want to do this if you were maxing out your CPU or disk IO or network capacity on a particular server.
Yes, I get the merit of load balancing. If I am load balancing over multiple VMs on multiple physical boxes, as soon as a host/VM is dead, the load balancer takes the server/VM out of the pool of where to direct clients. One request or so dropped. Very small downtime.
If a host dies but I have to wait for the VM to failover to a second node as part of a windows failover cluster, that takes much longer. The cluster has to realise the VM is down, then bring up and boot it on another host.
By having load balancing do this, as soon as the IIS server stops serving HTTP requests, the LB would take it out of the pool and traffic continues...
I also get the benefit that I can take a VM out of the pool to patch etc, without bringing the service down.
-
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Currently, if IIS website A is running on a VM on Host1, and that host dies, we're down. We have to turn the replica VM on which is on Host2. That's not automatic. By having that website on two VM's, one on Host 1 and one on Host 2, with a load balancer, the lb would stop directing traffic to the failed one and push everything to the live one. The service is better for customers...
I think you are confusing failover with load balancing. I totally see the benefit to having failover, I don't see any to load balancing. That's why I keep prying. I think that you are looking for the wrong solution. Not that the products aren't the same, but your needs and goals are different.
Load balancing is for when you are too big for one system to host. Until you are that big, load balancing is a negative.
Part of load balancing is to stop directing traffic to down instances right? If
Not really. That's failover. Loadbalancers do that, but it is NOT load balancing. It's a different action. What you want is failover WITHOUT load balancing.
-
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Wouldn't load balancing give us what we want here? Load balancer will direct traffic away from a down node. Yes, its not a 'failover', you're right... but customers are not affected...
Absolutely not. Pure load balancing would keep directing traffic to the dead node. You are mixing concepts together because people often use the same devices for both and have gotten stuck talking about the wrong one of the two.
-
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Currently, if IIS website A is running on a VM on Host1, and that host dies, we're down. We have to turn the replica VM on which is on Host2. That's not automatic. By having that website on two VM's, one on Host 1 and one on Host 2, with a load balancer, the lb would stop directing traffic to the failed one and push everything to the live one. The service is better for customers...
I think you are confusing failover with load balancing. I totally see the benefit to having failover, I don't see any to load balancing. That's why I keep prying. I think that you are looking for the wrong solution. Not that the products aren't the same, but your needs and goals are different.
Load balancing is for when you are too big for one system to host. Until you are that big, load balancing is a negative.
Part of load balancing is to stop directing traffic to down instances right? If
Not really. That's failover. Loadbalancers do that, but it is NOT load balancing. It's a different action. What you want is failover WITHOUT load balancing.
It is a different action, yes. Correct. It is however included by using a load balancer. So why wouldn't I use that for accomplishing this?
-
@Romo said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Load balancing lets you spread load over multiple servers. You would want to do this if you were maxing out your CPU or disk IO or network capacity on a particular server.
And to clarify, you would ONLY want to do this if that were true.
-
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Wouldn't load balancing give us what we want here? Load balancer will direct traffic away from a down node. Yes, its not a 'failover', you're right... but customers are not affected...
Absolutely not. Pure load balancing would keep directing traffic to the dead node. You are mixing concepts together because people often use the same devices for both and have gotten stuck talking about the wrong one of the two.
Ok, fair do's. So what would you suggest to use here?
-
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Currently, if IIS website A is running on a VM on Host1, and that host dies, we're down. We have to turn the replica VM on which is on Host2. That's not automatic. By having that website on two VM's, one on Host 1 and one on Host 2, with a load balancer, the lb would stop directing traffic to the failed one and push everything to the live one. The service is better for customers...
I think you are confusing failover with load balancing. I totally see the benefit to having failover, I don't see any to load balancing. That's why I keep prying. I think that you are looking for the wrong solution. Not that the products aren't the same, but your needs and goals are different.
Load balancing is for when you are too big for one system to host. Until you are that big, load balancing is a negative.
Part of load balancing is to stop directing traffic to down instances right? If
Not really. That's failover. Loadbalancers do that, but it is NOT load balancing. It's a different action. What you want is failover WITHOUT load balancing.
It is a different action, yes. Correct. It is however included by using a load balancer. So why wouldn't I use that for accomplishing this?
It is included in SOME load balancing. Only when the device in question is more than a load balancer. Why do you not use a car for watching television even when it includes seats? You are looking for the wrong action, you are really stuck on this. HA-Proxy or NGinx, which are proxies that do both failover and load balancing, will do what you want... because they have failover proxies NOT because they are load balancers, you want their load balancing component shut off! If you only wanted load balancing ,CloudFlare will do that for you.
-
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
Yes, I get the merit of load balancing. If I am load balancing over multiple VMs on multiple physical boxes, as soon as a host/VM is dead, the load balancer takes the server/VM out of the pool of where to direct clients. One request or so dropped. Very small downtime.
No, you are mixing the concepts back together. He just split them out for you. Failover does what you want, LB does not.
-
NGinx and HA-Proxy are fine tools for this, just DO NOT use them as load balancers. No reason for that complexity, it will have no benefits for you, but will have negatives.
-
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
NGinx and HA-Proxy are fine tools for this, just DO NOT use them as load balancers. No reason for that complexity, it will have no benefits for you, but will have negatives.
Yes, I see this now. Thank you. Wrong terminology from me. My goal then is to have multiple IIS instances running on different hardware (on VMs on different hardware), being routed to through a pair of (somethings?) which will stop routing to any of those sites that are down.
-
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
NGinx and HA-Proxy are fine tools for this, just DO NOT use them as load balancers. No reason for that complexity, it will have no benefits for you, but will have negatives.
Yes, I see this now. Thank you. Wrong terminology from me. My goal then is to have multiple IIS instances running on different hardware (on VMs on different hardware), being routed to through a pair of (somethings?) which will stop routing to any of those sites that are down.
Right, yes, and that's why HA-Proxy doesn't have Load Balancing in its name, but rather High Availability. Because failover is its primary use case.
-
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@Jimmy9008 said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
@scottalanmiller said in New Project - Thoughts? (CentOS, HAProxy, Load Balance)...:
NGinx and HA-Proxy are fine tools for this, just DO NOT use them as load balancers. No reason for that complexity, it will have no benefits for you, but will have negatives.
Yes, I see this now. Thank you. Wrong terminology from me. My goal then is to have multiple IIS instances running on different hardware (on VMs on different hardware), being routed to through a pair of (somethings?) which will stop routing to any of those sites that are down.
Right, yes, and that's why HA-Proxy doesn't have Load Balancing in its name, but rather High Availability. Because failover is its primary use case.
So NGINX over HAProxy? Or something else?
