@scottalanmiller some machine have complicated configuration that are IP-bounded but not netmask-bounded. The machines on the first subnet are already on DHCP but of course with reservation (that I will maintain unaltered).

@scottalanmiller said in Any reason to avoid /16 in 2017?:

@francesco-provino said in Any reason to avoid /16 in 2017?:

@travisdh1 said in Any reason to avoid /16 in 2017?:

@francesco-provino said in Any reason to avoid /16 in 2017?:

@scottalanmiller said in Any reason to avoid /16 in 2017?:

@wrx7m said in Any reason to avoid /16 in 2017?:

A /16 seems pretty drastic when coming from a /24

Yeah, /16 is too large to actually use. Commonly /23 and /22 are used, they are no problem. In reality, /21 is perfectly fine. Even a /20 is pretty good. But once you start getting into the /19 and larger, you are just getting to some impractically large scales. A /16 is 16,000% larger than a /20, which is generally considered the largest that you can practically use.

I know this is the common sense, but… what will be the issue? I will have just 300-350 allocated IP, it's just a matter of convenience to include both the X.X.0.0 and the X.X.120.0 range in one big subnet.

I know it doesn't matter much, but the AWS VPC subnet is /16 by default :D.

After a certain point, broadcasts overwhelm actual network traffic. That's really the only thing I know that limits the size of a single network.

Ok, but I think the broadcast traffic depends only on the number of hosts in the subnet. I wouldn't put more than ~500 active IPs in this subnet, ever.

Then why go to something so absurdly large instead of just something 2-4x larger than your maximum possible usage?

Again, because there are two production networks that I want to merge together, one is X.X.0.0/24 and the other is X.X.120.0/24. It's hard to rebuild every static-ip-bounded configuration in our small maintainance window, so I plan to change it piece by piece to DHCP reservation, but I cannot do it at one time.

But, really, what's the problem with the "wasted" space? Is there something intrinsically dangerous or heavy to compute with a /16 network for modern equipment?

@travisdh1 said in Any reason to avoid /16 in 2017?:

@francesco-provino said in Any reason to avoid /16 in 2017?:

@scottalanmiller said in Any reason to avoid /16 in 2017?:

@wrx7m said in Any reason to avoid /16 in 2017?:

A /16 seems pretty drastic when coming from a /24

Yeah, /16 is too large to actually use. Commonly /23 and /22 are used, they are no problem. In reality, /21 is perfectly fine. Even a /20 is pretty good. But once you start getting into the /19 and larger, you are just getting to some impractically large scales. A /16 is 16,000% larger than a /20, which is generally considered the largest that you can practically use.

I know this is the common sense, but… what will be the issue? I will have just 300-350 allocated IP, it's just a matter of convenience to include both the X.X.0.0 and the X.X.120.0 range in one big subnet.

I know it doesn't matter much, but the AWS VPC subnet is /16 by default :D.

After a certain point, broadcasts overwhelm actual network traffic. That's really the only thing I know that limits the size of a single network.

Ok, but I think the broadcast traffic depends only on the number of hosts in the subnet. I wouldn't put more than ~500 active IPs in this subnet, ever.

@scottalanmiller said in Any reason to avoid /16 in 2017?:

@wrx7m said in Any reason to avoid /16 in 2017?:

A /16 seems pretty drastic when coming from a /24

Yeah, /16 is too large to actually use. Commonly /23 and /22 are used, they are no problem. In reality, /21 is perfectly fine. Even a /20 is pretty good. But once you start getting into the /19 and larger, you are just getting to some impractically large scales. A /16 is 16,000% larger than a /20, which is generally considered the largest that you can practically use.

I know this is the common sense, but… what will be the issue? I will have just 300-350 allocated IP, it's just a matter of convenience to include both the X.X.0.0 and the X.X.120.0 range in one big subnet.

I know it doesn't matter much, but the AWS VPC subnet is /16 by default :D.

@wrx7m said in Any reason to avoid /16 in 2017?:

One thing to note is that if you slowly migrate and you have systems that come from beyond the /24, you will get a lot of traffic looking for the gateway to get to that other subnet. I went from a /24 to a /22 several years ago and am still good to go.

Whichever way you go, make sure that you adjust your DHCP lease times to a very brief time so that they will get the new IP scheme sooner than later.

Edit: adjust the DHCP lease times far enough in advance so it will make a difference. After they all have new leases with the new scheme, you can lengthen the lease time.

Very helpful tips, thank you!

Hi everyone, I was just thinking about merging two /24 subnet that are like X.X.0.0 and X.X.120.0 (running out of IPs)… some apps that run on that contains weird static-ip configuration (like SAP B1 services with hardcoded IP, CCTV ecc), so I cannot easily reconfigure everything from scratch with DHCP in a bigger subnet.

So, I was thinking about just use X.X.0.0/16 and slowly migrate the static stuff to the new mask (and migrate to DHCP with reservation, of course), but I've read too much good-old advices about "too big domain", "too much broadcast traffic" ecc to use the /16 lightly.

Do you know about any real issue about using /16 in a LAN? Here it's all modern ubiquiti switches, AP and router. The other machines are fairly new, too.

I run two VMware Horizon environments for two companies. My advice is… avoid VDI. At any cost.
The only scenario in which it makes sense is a security-tight environment (banks, military, government, strong NDA etc.).
For ANY other case, just use proper management tools, desktop imaging etc.

The VDI is a land full of pain, believe me.

Just buy an iPad pro with keyboard or any macbook 12.
Develop on a hosted machine, in your company colo or any cloud you like.

It's cheaper and more reliable than any of the other alternatives, I've already been on that path.

Going deeper: XPS 13 has coil whine and horrible webcam position; UX390UA runs too hot, especially with Linux.

@ghani why not? Of course, a replicated vCenter or a clustered one is the way to go.

Why don’t add both prod and DR hosts to one vcenter?

@dashrender said in Finding the Best Open Source Email Solutions:

Lol now Scott seems to be advocating for self hosted email.

I agree with you, what an u-turn @scottalanmiller !

@coliver said in Upgrading from ESXi 5.5 to ESXi 6.5 . what are the steps need to follow ?:

@francesco-provino said in Upgrading from ESXi 5.5 to ESXi 6.5 . what are the steps need to follow ?:

@coliver said in Upgrading from ESXi 5.5 to ESXi 6.5 . what are the steps need to follow ?:

@francesco-provino said in Upgrading from ESXi 5.5 to ESXi 6.5 . what are the steps need to follow ?:

The VCSA installer is incredibly fragile and bugged. Of course it doesn't makes any sense to deploy vCenter on windows as of today.

VCSA is an appliance isn't it? There shouldn't really be an installer.

I totally agree with you. But there is an installer with a two-stage deployment that fail spectacularly in 95% of the deployments, mainly due to DNS or NTP issue. The VMware KB is full of workarounds like "put this MANUALLY on the host file".

Ouch. Wasn't aware of that. I know two years ago when we went to deploy it we had nothing but issues getting it working and the interface didn't work at all. Hopefully the modern appliance is much better.

The appliance now is usable, but the installer is completely useless.

@coliver said in Upgrading from ESXi 5.5 to ESXi 6.5 . what are the steps need to follow ?:

@francesco-provino said in Upgrading from ESXi 5.5 to ESXi 6.5 . what are the steps need to follow ?:

The VCSA installer is incredibly fragile and bugged. Of course it doesn't makes any sense to deploy vCenter on windows as of today.

VCSA is an appliance isn't it? There shouldn't really be an installer.

I totally agree with you. But there is an installer with a two-stage deployment that fail spectacularly in 95% of the deployments, mainly due to DNS or NTP issue. The VMware KB is full of workarounds like "put this MANUALLY on the host file".

The VCSA installer is incredibly fragile and bugged. Of course it doesn't makes any sense to deploy vCenter on windows as of today.

@ghani said in Upgrading from ESXi 5.5 to ESXi 6.5 . what are the steps need to follow ?:

@francesco-provino said in Upgrading from ESXi 5.5 to ESXi 6.5 . what are the steps need to follow ?:

vSphere upgrade path on vmware si

i searched, but that they do not provide sorted VMware upgradation pre-requistics in esxi host level, vcenter level, VM level, storage level, network level ???

Of course they require SOME of that: https://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=2147289 .

More general compatibility matrix: https://www.vmware.com/resources/compatibility/sim/interop_matrix.php .

First, read the vSphere upgrade path on vmware site (just google it), after that…

It looks nice on first sight. Any other open source competitor? I mean, oss with HTML5 interface.

I've just discovered this OSS inventory tool by digital ocean: https://github.com/digitalocean/netbox . Does anyone has any experience with that? AFAIK DO makes great stuff and docs, hope this is not an exception!

@guyinpv +1 for Trello

@jt1001001 trello.