|
PIEBALDconsult wrote: waiting patiently for me to apply updates
Probably an important distinction:
I'm not talking about waiting for me to install updates; I do have it install them (since that's something that can take a while, but can run in the background while I keep working).
I just want it to wait for me after that to give the okay to reboot. That's how I've always done it. But in this particular case, I imagine, the prompt comes up while I'm away, it gets no response, then reboots on its own overnight. Server 2012 R2 (and probably others) waited indefinitely. 2022 won't even wait for 24 hours, whether you've seen the prompt or not.
|
|
|
|
|
dandy72 wrote: Surely server admins aren't putting up with this.
Not sure what you mean.
Standard large distributed system architecture design would be to expect servers to reboot, fail, and even just disappear (taken down and not restored.)
As an example AWS SLA is 99.99% per month. So it will fail. You only get back (money) for the time it was down if it was down for more than that. And it is generally up to you to figure it out and prove it.
dandy72 wrote: so a server OS can now reboot if it just feels like it.
I believe one can turn patching off entirely.
That however only prevents reboots due to a patch. Restarts for other reasons are possible. Some that I can think of
- Manual reboot request
- Perhaps detected error. So OS and/or hardware detected problem and restarted.
- Power problem.
dandy72 wrote: a server OS can now reboot if it just feels like it.
Perhaps not applicable to you but at least AWS will force updates for certain cases. They give notice but if the user has not updated the system by the specified date they will just do it.
|
|
|
|
|
dude, some of us may run "server" OS, but I suspect we're talking about a machine almost certainly used in development. I do expect interruptions, it's why I have a UPS in my office. The context is MS forcibly rebooting machines because they are just stupid and have their heads where the sun doesn't shine. I'm being polite. Ask me how I really feel. There is not another OS in the world that forces updates/reboots like this. It's just stupid.
Customers I work with have some sort of enterprise version where the forced reboot is clearly disabled. I've logged into these machines for months and seen "you need to reboot" popups. I'm not the sysadmin, so it's not up to me to reboot things.
Charlie Gilley
“They who can give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety.” BF, 1759
Has never been more appropriate.
|
|
|
|
|
charlieg wrote: The context is MS forcibly rebooting machines
The OP does not state that they found the cause of the reboot. As I noted in my response there are a number of possible causes.
charlieg wrote: where the forced reboot is clearly disabled.
Which applies to one situation.
|
|
|
|
|
jschell wrote: Standard large distributed system architecture design would be to expect servers to reboot, fail, and even just disappear (taken down and not restored.) We are not talking about a unforeseen failure that crashes a server.
We are talking about an outside third party intentionally rebooting a server at its whim, completely ignoring the needs of the owner of that server.
For former is an "act of god". The latter is an "act of Microsoft".
jschell wrote: Perhaps not applicable to you but at least AWS will force updates for certain cases. They give notice but if the user has not updated the system by the specified date they will just do it. Emphasis mine.
This is a totally different scenario, as the owner of the servers (AWS) gives notice to the tenants that maintenance is required, and acts when a deadline has been reached.
Similarly, all computers and major systems at my employer undergo periodic maintenance. The respective owners give users (tenants) sufficient warning ahead of time, and any who fail to heed the warning, get what they get.
|
|
|
|
|
BryanFazekas wrote: We are not talking about a unforeseen failure that crashes a server.
We are talking about an outside third party intentionally rebooting a server at its whim, completely ignoring the needs of the owner of that server.
This. So much this.
BryanFazekas wrote: For former is an "act of god". The latter is an "act of Microsoft".
I think the fundamental problem here is that according to Microsoft, there is no difference between the two.
|
|
|
|
|
BryanFazekas wrote: We are talking about an outside third party intentionally rebooting a server at its whim
The OP did not state that.
BryanFazekas wrote: (AWS) gives notice to the tenants that maintenance is required, and acts when a deadline has been reached.
And yet in my experience people are still surprised when it happens. The notice was given but they did not see it so they do not understand why it happened.
BryanFazekas wrote: Similarly, all computers and major systems at my employer undergo periodic maintenance.
In my experience I have had to debug production server problems because someone did an update without telling anyone. Sometimes without even documenting it. Even when documented often without documenting what the exact update did.
And then when I am tasked with determining the problem, no one on the call is even aware that an update happened.
|
|
|
|
|
jschell wrote: The OP did not state that.
It was not stated clearly, but who else can reboot his server after Patch Tuesday other than Microsoft? The fact that with Win10 Microsoft DOES force reboots following an update, plus the context of the post indicate it's MS.
|
|
|
|
|
BryanFazekas wrote: who else can reboot his server after Patch Tuesday other than Microsoft
There are multiple reasons a server can reboot. I provided a list in another reply.
And any of those others would result in the same problem, because the person is expecting the server to never do that.
|
|
|
|
|
dandy72 wrote: Surely server admins aren't putting up with this. In the book by Raymond Chen, The Old New Thing (well, since the book is a selection of his blog posts, you can find it on the net as well), he tells a story about a server that had a small memory leak causing it to crash at intervals of a few days. The server had to be available 24/7, and debugging it while it was operational was not possible.
So they installed two servers, with load balancing software. When one of them was getting close to memory saturation, they trimmed the load balancer to send all new tasks to the other machine, let those jobs already on the first one run to completion, and then they could start the analysis of the memory structures. The other machine was filling up, so they had to reboot the first one, with a clean heap, and turn the load balancer the other way, to send all new jobs to the first machine, and do further debugging on the second one.
As long as they were debugging, and afterwards, fixing the software, they kept the load balancer switching jobs to alternating servers while the opposite one was rebooted. Once the software was fixed, there was no more need for two servers; the original one had alone much more capacity than needed, so the other one could be returned.
Stories like this are what makes "The Old New Thing" really worth reading. There are several of them.
(The book is from 2006, so it is somewhat 'historical' - and yet it is worth reading. So is his blog, which is updated several times a week, almost daily. For those unaware of it: Raymond Chen: The Old New Thing[^])
|
|
|
|
|
that's a cool hack/workaround.
CI/CD = Continuous Impediment/Continuous Despair
|
|
|
|
|
it's the same old elephanting Microsoft bullcrap. Allegedly, there is a way to stop it, but there is a special place in hell for the manager/team or whatever that forces this crap.
Charlie Gilley
“They who can give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety.” BF, 1759
Has never been more appropriate.
|
|
|
|
|
If you keep up that style, you run the risk of all your posts also being classified as "bullcrap" and "bologna".
|
|
|
|
|
But as the OP asked, is it appropriate to just reboot servers "because"? And the answer is no. Never. And it got so bad with these idiots they had to put enterprise hooks in to turn this off.
One writes software to be stable. Microsoft has deliberately engineered a sporadic reboot that none of my customers would ever tolerate. It's criminal.
Let me give you one simple example. I need to be able to run soak tests for weeks/months at a time. I have multiple UPS', I am very careful with my hardware, etc. I've developed under openVMS, Solaris, HPUX, IBM AIX, Linux and Microsoft. The only company and OS that forces reboots is Microsoft.
So, I'm not sure what the point of your comment is. The OP asked a simple question, and the answer is no. No OS should ever just reboot itself. So, I picked bologna to avoid profanity. You really don't want me speaking what I'm really thinking.
Charlie Gilley
“They who can give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety.” BF, 1759
Has never been more appropriate.
|
|
|
|
|
charlieg wrote: So, I'm not sure what the point of your comment is. Just ignore him. Trust me. He's got issues man. There was nothing wrong with your post.
Jeremy Falcon
|
|
|
|
|
I stop my car to refuel it. Has never caused me any problems.
|
|
|
|
|
So if Boeing and Airbus randomly shutdown the aircraft they sold to airlines you'd be OK with that too?
|
|
|
|
|
But here is the deal: "I stop my car to refuel it."
You do that. Now imagine going down the road and having your car reboot itself... without your permission. I'm not quite sure what the argument is. Tell you what. Go into work and start unplugging peoples' computers and tell them it's for their own good. please post copies of your medical bills .
Charlie Gilley
“They who can give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety.” BF, 1759
Has never been more appropriate.
|
|
|
|
|
But as the OP asked, is it appropriate to just reboot servers "because"?
Yes, it is. The world is full of Tao practitioners.
|
|
|
|
|
You've never been employed as a server admin, have you?
Or if you have...I'm guessing exactly once and never again.
|
|
|
|
|
|
Peter Adam wrote: Life is too short to wait for everyone to be in the right moment.
Sometimes, that's just not your call. A buddy of mine also worked for a large company; he was responsible for getting things patched, and sometimes, an upgrade and subsequent reboot required a sign-off from 8 different people agreeing on a time and date.
Peter Adam wrote: Maybe they are running critical infrastructure
They too
Sure, some admins are downright irresponsible and never apply patches. For a decade, exactly as one of your links points to.
But ultimately, patching/rebooting is the admin's responsibility; whoever makes a patch can't make those decisions on anyone else's behalf. That's what I'll object to and I'll fight this tooth and nail.
|
|
|
|
|
charlieg wrote: No OS should ever just reboot itself
I can remember my surprise when running on a Solaris system when it would do a cold boot when I attempted to run Netscape. Every single time.
Pretty sure the OS and/or the hardware was in fact in charge of doing the cold boot.
Myself I design for failure. Boxes will go down. Doesn't matter why.
|
|
|
|
|
that's a bug and I think you are going way, way back . It's annoying but completely different from a corporate policy that says FU.
Charlie Gilley
“They who can give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety.” BF, 1759
Has never been more appropriate.
|
|
|
|
|
charlieg wrote: but completely different from a corporate policy that says FU.
Not at all.
Microsoft sells to consumers. It is nice and all that there is also a large scale professional server usage but that is not the primary market.
And as it has been proven with all OSes, not just Microsoft, people ignore updates. So now Microsoft is just forcing it.
Just to enforce that - linux has had multiple security bugs with patches already provided but which professionals failed to install until criminals started using them. Some of them were known for years.
Perhaps those same people that think the server should be up for years.
(I have worked for multiple companies where managing even certs is a problem. They know the cert will expire on a certain date yet every time there is a mad scramble the morning after it expires to figure out what is going on.)
|
|
|
|