Enterprise Microsoft Q&A: Hello for Buttprint

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > Enterprise Microsoft Q&A: Hello for Buttprint

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

we pretty much let users do whatever they want and then hire twice as many techs as we would otherwise need to fix everything

# ¿ Jul 13, 2010 14:30

Adbot: ADBOT LOVES YOU

# ¿ May 5, 2024 19:17

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

BooDaa posted:

So is SCCM going to give me anything that I'm not getting in WSUS 3.2 as far as MS updates are concerned?

Software updates are only a tiny piece of the security picture -- it's just as important to have visibility into your systems and know that their configurations are in compliance with org policy. But if all you need or care about is a single update profile to push out to your hosts, SCCM is probably overkill and wasted money for you and you can probably do most of what you need between WSUS and good use of GPOs.

# ¿ Jul 21, 2010 20:18

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

There's a huge number of multi-tenant management products out there for managed service providers. Kaseya, N-Able, Level Platforms and ManageEngine are the most popular that I'm aware of, though I haven't used any personally, being a Linux admin that does not work for an MSP.

# ¿ Jul 26, 2010 15:36

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Long-shot, but is anyone here using Graphite to aggregate metrics on Windows systems? What are you using to feed data into it?

# ¿ Nov 10, 2011 21:20

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

evil_bunnY posted:

Speaking of metrics, what do you guys use for cross platform metrics acquisition?

We've been using Nagios with the PNP4Nagios addon to translate plugin performance data to RRDTool graphs automatically, but we're trying to ditch it because it's impossible to do any meaningful aggregation with it. Unfortunately, this seems to be the least worst option until you get way up into the more expensive commercial monitoring options.

Edit: I'm writing a metrics daemon called winmetricsd using .NET 4 and the Reactive Extensions to feed WMI perf counter data asynchronously into Graphite. Hopefully I'll have something usable in a few weeks. I'm targeting Graphite initially by implementing the collectd network protocol, but I'm planning on making a future release plugin-based so it can target OpenTSDB and other storage backends.

Vulture Culture fucked around with this message at 13:37 on Nov 11, 2011

# ¿ Nov 11, 2011 13:12

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

evil_bunnY posted:

Sounds baller.
Q: why not just build a Windows WMI collector for OpenTSDB?

When you're collecting a lot of metrics at a rapid sampling interval, WMI actually has a lot of overhead, especially when you're taking metrics with a discrete sampling interval like CPU usage. It becomes very difficult to do things like send 100 different metrics every 60 seconds when you're reliant on the ability of WMI to keep pace with the collector; remote WMI doesn't play nice with lots of concurrent asynchronous calls. The registry interface, by comparison, is much leaner than WMI/WBEM, so it should have much less of a performance impact. Plus, push is almost always preferable to pull. You can fire and forget and not worry about latency from the remote end slowing anything down, you just drop some metric data if the monitoring host is overtaxed.

Agents suck, but setting up WMI in a mixed environment and getting authentication to actually work right from your collector host sucks more, especially when a team's understanding of Kerberos is "the thing that makes AD logins work." This keeps Kerberos out of the mix entirely.

My last rationalization for making it an agent is that the counter you want to use sometimes varies. For example, the CPU and memory usage counters are pretty much worthless on virtualized systems. VMware Tools, however, provides additional performance counters that do the right thing and display the correct information.

I like it when things are easy, even if it means more work up front.

In the long run, I wouldn't mind if someone extended ESxSNMP to support WMI in addition to SNMP, but that probably won't happen anytime soon.

Vulture Culture fucked around with this message at 15:17 on Nov 11, 2011

# ¿ Nov 11, 2011 15:11

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Bitch Stewie posted:

So, how are you all provisioning file volumes on your file servers?

Our SAN has a single big virtualised storage pool so I don't have to deal with RAID groups at SAN level, I just get an Xtb storage pool to play with.

I currently provision multiple volumes of 2tb each to the file server, but of course you end up with some areas data on volumes with no free space and other volumes have lots of free space.

A single big volume doesn't seem sensible as it makes things like backup and chkdsk "fun".

We avoid this problem by not running NTFS file servers. There's too many good storage platforms out there to waste time trying to roll our own and have them subject to these sorts of problems.

# ¿ Dec 21, 2011 05:59

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Nebulis01 posted:

What the hell does that have to do with NTFS? NTFS supports volume size of 256TB using a GPT disk. As long as your server is windows 2003 SP1 and above this hasn't been an issue in years.

NTFS lacks end-to-end integrity checking and other availability features that are implemented in higher-end filesystems like ZFS, GPFS2, OneFS and (ugh) btrfs that are built to scale. As a result, it's very poor at detecting corruption while the filesystem is online. That means that when the poo poo hits the fan, NTFS has to take the disk offline to run chkdsk. Have you ever tried to run chkdsk on a disk that's more than a few terabytes in size? It typically takes weeks where most proper filesystems take literally zero time because it's something they just constantly do in the background.

No offense, but have you ever run significant storage in production?

# ¿ Dec 21, 2011 21:56

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

FISHMANPET posted:

Yeah, I'm aware of all that. We're currently in the midst of trying to consolidate storage into some kind of SAN and we were planning on just sharing iSCSI to all the machines because we have an irrational hatred of NAS stuff.

vv Welp, my job is weird.

NTFS isn't a clustered filesystem. You literally will be constantly corrupting the entire filesystem every time another client updates anything.

# ¿ Dec 24, 2011 05:43

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

FISHMANPET posted:

Whoops, that's not what I mean. I mean all our storage would be provisioned to servers that would then share that space out to clients, either via NFS or SMB, rather than using the device as a NAS.

A lot of people who don't need gigantic monolithic filesystems take this approach with DFS or something similar, because the alternatives are all pretty gross.

# ¿ Dec 24, 2011 14:23

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Wicaeed posted:

Isn't that what File Screens are for in the first place? Prevent users from saving .wav .mp3 .flac .mp4 files, voila!

Unfortunately, several of our worst offenders also have a terabyte of recordings of bird songs that are completely legitimate and must be retained to comply with a federal grant.

# ¿ Dec 27, 2011 19:56

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

The Macaroni posted:

Not sure if this is the right thread for this--please tell me to go away if I'm wrong.

At my new job, we're running Windows 7 and Internet Explorer 8 throughout the enterprise. No doubt you're familiar with the obnoxious IE mixed content error that requires you to click "No" to load a webpage which calls both secure and non-secure content. I know how to change the security settings on an individual computer to make this go away. But I want to ask our IT folks to change that setting for all users--I expect they'll tell me to deal with users on a case-by-case basis, but that doesn't seem efficient to me.

My question: in 2013, is there a good reason to continue to have that setting prompt the user to continue? Is there a good reason not to change that setting for all users? I tried Google and it just pulls up explanations on disabling this on an individual computer.

On a side note, get ready for a lot of websites to drop IE8 support within the next couple of months; jQuery 2.0 has stopped supporting IE8 and below.

Hooray for your company! :downs:

# ¿ Apr 23, 2013 04:04

Adbot: ADBOT LOVES YOU

# ¿ May 5, 2024 19:17

Vulture Culture: Jul 14, 2003; I was never enjoying it. I only eat it for the nutrients.

Are there any good free replacements for Steady State floating around yet? I played around with Reboot Restore Rx and it works as advertised, but I'd really like something with some kind of CLI for scripting updates to the baseline.

# ¿ Dec 15, 2015 21:02

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > Enterprise Microsoft Q&A: Hello for Buttprint