The power issues I'm having are with a Macbook, where it won't go to sleep when you close the lid.

The link I posted was about how Microsoft basically fired their entire testing staff, instead leaving testing now up to the end user. While I'm not a huge OS X fan, Windows 10 is still a rubbish product, especially the forced updates which break more things than they fix. Office 365 is the same story. Makes me glad to run a version or two behind on most Microsoft stuff and 100% of our back-office is NOT running anything MS.

The problem is Microsoft's hiring practices. Anyone diligent enough can be taught to properly test patches (although whether they will or not is another story). But Microsoft hires a lot of people who should not be interacting with clients (at least not yet) I work for a Microsoft Platinum partner and when we need to engage Microsoft, we start to shake our collective heads as it's usually a crapshoot as to how knowledgeable a person we get. Often they're newer support staff that don't know much and need to pull in a level two guy.

That said, when the MS product works, it generally works well for us. It's when the client gets creative that we start to have issues (eg, federated authentication works fine for their mobile devices and current MDM product. But then a c-suite talks to a vendor and wants to flip over to the new MDM product without evaluating whether it's suitable or not for them... the sales guy oversold the software... and voila, nothing works properly until we flip it back)

If you’ve disabled Fast Shutdown, the following should help:” with the power button action, fairly easy to set

https://www.tenforums.com/tutorials/...dows-10-a.html

Thanks. I should have been clearer. That is exactly what I have done.

I still have an MS support ticket open from 2008. Even the latest versions of MS Office still have the same bug... I gave up on it getting fixed.

My gripe is when I find a work-around and get things going again, then MS "fixes" it in a patch and things don't work again, including my hack.

That's an ouch. Never had anything go that long.


Trust me. Had an issue that happened with a few clients where if a server was configured in a certain way, after about a week or so, a server would go offline due to resource exhaustion. There were two fixes. First was a reboot (which you can't keep doing weekly in a production environment). The second fix was to disable some settings. Unfortunately MS's patches would randomly re-enable some of those settings (there was no rhyme or reason as to when it would happen, it's just it would be part of the patch). The worse part was that it wouldn't always show up in the non-prod environments but randomly in production. Come to think of it, I don't think the issue was ever properly resolved... and this goes back to 2010 or so...

Actually.... our main office used to reboot their servers on Fridays at 12 noon. All of them. I thought it was ridiculous. Fortunately we only used their systems for time & billing. As you might expect, there were quite a few times when everyone got sent home by 2pm when a server decided not to come back up and it wasn't going to be an easy fix. They didn't have any redundancy either. ::facepalm::

Oddly enough, they split the company and we inherited the time & billing system, but no hardware to run it on. It was a mess. the DB and tables all were fragmented 85-98%. Nothing was documented. I cleaned it up and shoehorned it into a Win10 (updates disabled) XCP-NG VM running MS SQL Express and it's been well-behaved. Not a single reboot in 20 months. Performance, even on a circ-2007 Dell 2950, VM'd, is light years faster than when they had it. Snappy, actually. Even the worst report takes <2 seconds to complete. I'm not sure if it's incompetence or all of the overhead of a full MS server + full MS SQL, although I have my suspicions.

There was probably a good reason why nobody changed a hugely inefficient system as you describe.

Many staff had obviously enjoyed the benefits of regular Friday afternoons off and anybody proposing to end it would have got short shrift!

I don't think I've encountered a full SQL instance that couldn't use a reboot every month. One client was funny though. They ran a three node SQL cluster. Every 2-3 months a random node would go offline. The DBA who previously managed their SQL for them said that was normal. I just happened to be there to drop off some goodies (as they were relatively new) and heard this. No patches for at least a year, transaction logs were full (not sure for how long) and they had just rebooted a downed node. Had one of my DBAs purge the logs and set up resourcing restrictions (and other cleanup tasks). Got them to try having my team maintain their cluster for two months (eg, patching with failover, purging logs and other things).... it's been a year and a half since and haven't heard of any issues since.

I've been meaning to learn SQL myself (at least some basic maintenance stuff anyway) but work has been busy and I've been learning other things at the same time... so yeah, one day.

Actually that brings up one thing. Which would you recommend I learn first (need to learn both). XCP-NG or Proxmox? Curious... a lot of the vloggers are flogging proxmox with one or two promoting XCP-NG so interested in finding out what non-vloggers would suggest.

I've got 1 server (effectively a NAS) and the backbone between the old part of the house and the addition on 10GbE. The former is "because I can" although it is nice to be able to saturate multiple 1Gb streams going out if multiple people are watching videos (not that any of the video streams I have are anywhere close, but VLC on Windows and Kodi on Android/Android TV both seem to buffer extraordinarily aggressively when they're using windows-style file sharing to get at the videos.)

The backbone was definitely worth it - saved me running a lot of copper between the sides of the house, and over a ~100 foot run it may not have saved me money but it was definitely worth it for spending less than 1/5th the time in the crawl spaces. Once I realize how much lighter and easier pre-terminated fiber is for very long runs, I never want to go back for longer runs and in-wall use.

What kind of budget did you work with? I'd like to redo the computers I have in the house to reduce the load on the wifi network and if I can speed things up for a reasonable price, I'd consider it. But the prices I see even for 2.5 is not cheap.

Honestly, since I did the conversion and DB optimization, I've not done a single thing to the DB other than backing it up every other day. That was Feb 2020. It just runs, so I just leave it alone. Now that we're discussing it, I'll probably poke at it tomorrow and see how it's doing. No user complaints whatsoever.

I've never seen Proxmox in a production environment. It has a lot of nice features, especially being able to run ZFS natively, and having the console built in, but that's about it. Citrix XenSever (and XCP-NG) are found in production environments. They're meant for clustering and all sorts of other goodness. DO compile a XenOrchestra instance from source. The features are amazing. Live Migration's fantastic, as it being able to dump the whole VM as a single file. XCP-NG + TrueNAS is a beautiful combo if you've got the resources to put it together.

In general, gigabit out to individual machines is still going to substantially outperform wifi - 10GbE is best for switch-to-switch, or really high-traffic machines like a NAS/server or things in a home lab.

I put this together in 2016, and spent about a little over $1000 total for two managed 20-port switches (D-link DGS-1510-20, not still available new except as remaining stock as best I can tell), each with 18 gigabit ports (16 copper, 2 fiber SFP) and 2 SFP+ 10 GBE (about $750 for both back then), plus a NIC for my server (2-port SFP+ Intel) and about $100 for SFP modules (cheaper now.) 100' fibers were cheaper then, but I bought them locally so I don't remember how much - probably about $50 for both of them

To do it the standard "everything goes back to 1 closet, and a single big managed switch" would have been probably $150+ and maybe another $60+ for a smaller 2nd spool just for the extra copper to do 10-12 runs under the house, plus a patch panel, plus a lot more wear and tear on me. :) A good 24+ port managed switch wouldn't have been much cheaper than one of the 16+uplinks, so I'm already at half the cost even before figuring out how to uplink the server properly (port bonding on 1GbE doesn't work all that well.)

It would have been quite a bit cheaper in late 2019, before supply chain shocks.

There are still cheaper options today, although not as many readily available: Microtik is really cheap for 10GbE: a 2-uplink model https://amzn.to/3oXnJ74 or https://amzn.to/3avkyeA for a 4-port 10GbE model (you can use the gigabit port as either a dedicated management port or a downlink on the same bridge) or https://amzn.to/2YDcwxX for an 8-port which would be a good backbone switch. These look interesting as well - https://amzn.to/3iT51tx for a cheaper uplink model

For fiber connections, you'll need something like this https://amzn.to/2YKGVKG for either end, or something like https://amzn.to/3iST8E5 but they only are good for lab. I thought they went to 10M but the longest I can find right now is 7.

10GbE copper switches are much more expensive, last I looked and on a cursory look on Amazon, but I may be missing something.

I use the 2-port version of this NIC: https://amzn.to/3awgf2q - there are cheaper ones, but the Intel X520 drivers are really stable. Unlike the switches, copper 10GbE NICs are often a little cheaper.

It depends on how active the DBs are. If you're talking about an SQL instance supporting a financial institution or a large retail chain, there's lots of work to be done. But if you're talking about a medium-sized consulting company, you can go a long time without touching it. But in a previous life I was at a large FI and they tend to be risk adverse so patching and active maintenance is important.

I was thinking XCP-ng, but a lot of people say go proxmox...Coming from a vmware world, a lot of the features in XCP-ng would be nice, Eventually I'll have the hardware to have both in my homelab, but until then, will have to figure this out. Thanks!

That's definitely going to cost a bit more here. I'm sure if I search hard, I could find great deals up here, but I find for a lot of hardware, Canadian prices are not on par with places in the US. But at least with what you've laid out, I have an idea of what to use. Microtik has come up several times as affordable (not here). For the most part, I can deal with 1GbE stuff, but once in a while, I could use faster when doing buildouts for clients... Also dealing with a dearth of wifi signals in the neighbourhood... so what I can take down I'd prefer to just to speed things up. Thanks! Now at least I know what to expect.