Never update, never reboot. Clearly the safest method. Tried and true.
Found the debian user!
Never touch a running system
Until you have a inviting hole in your systemNevertheless, I’m panicking every time I update my sever infrastructure…
when it was the wrong server and you’re hoping it comes back up before 5 minutes and nagios starts sending alerts
If a tree falls in the woods…
I install molly-guard on important machines for this reason. So fast to do a reboot on the wrong ssh session
Just had to restart our main MySQL instance today. Had to do it at 6am since that’s the lowest traffic point, and boy howdy this resonates.
2 solid minutes of the stack throwing 500 errors until the db was back up.
If you have the bandwidth… it is absolutely worth it to invest in a maintenance mode for your system, just check some flat file on disk for a flag before loading up a router or anything and then, if it’s engaged, just send back a static html file with ye olde “under construction” picture.
Bonus points if your static site sends a 503 with a retry after header.
That’s not really… possible at this point. We have thousands of customers (some very large ones, like A——n and G—-e and Wal___t) with tens or hundreds of millions of users, and even at lowest traffic periods do 60k+ queries per second.
This is the same MySQL instance I wrote about a while ago that hit the 16TiB table size limit (due to ext4 file system limitations) and caused a massive outage; worst I’ve been involved in during my 26 year career.
Every day I am shocked at our scale, considering my company is only like 90 engineers.
Is that the same database my user couldn’t connect to today?
When you make a potentially system breaking change and forgot to make a snapshot of the VM beforehand…
There’s always backups… Right?
… Right?
oh there is. from 3 years ago, and some
And no one ever tried to restore it.
Happened to me as well, after a year I learned incremental DB backups were wrongly offset by GMT diff, so we were losing hours every time. Fun.
Luckily we never needed them.
And now we have Postgres with WAL archiving and I sleep so much better.
Someone set up a script to automatically create daily backups to tape. Unfortunately, it’s still the first tape that was put in there 3.5 years ago, every backup since that one filled up failed. It might as well have failed silently because everyone who received the email with the error message filtered them to a folder they generally ignored.
I have more than once typed shutdown instead of reboot when working on a remote machine… always fun
Make an alias for Ehen you type shutdown it dies restart and if you want to shutdown make an alias that goes like
Yesireallywanttoshutdown
Networking, we had a remote office in Europe (I’m in the US) and wanted to reset a phone. Phone was on port 10 of the Cisco switch, port 1 went to the firewall (not my design, already in place).
Helping my coworker, I tell her to shut port 10.
Shut port 1, enter.
Ok… office is offline and on another continent…
i have the horrible habit of using shutdown now because of my personal computers. a lot more fun.
Not sure if this will help you, but I always do shutdown and then think about whether I want to do -r or -h. I’m sure it won’t help 🙂
Initializing VPC…
Configuring VPC…
Constructing VPC…
Planning VPC…
VPC Configuration…
Step (31/12)…
Spooling up VPC…
VPC Configuration Finished…
Beginning Declaration of VPC…
Declaring Configuration of VPC…
Submitting Paperwork for VPC Registration with IANA…
Redefining Port 22 for official use as our private VPC…
Recompiling OpenSSH to use Port 125…
Resetting all open SSH connections…
Your VPC declaration has been configured!
Initializing Declared VPC…
Y’all need high availability in your lives.
Tbh there is nothing more taxing on my mental health than doing maintenance on our production servers.
I work with IBM i/AS400 servers and those are not exactly the quickest thing to “reboot” (technically an IPL). Especially the old ones. I have access to the HMC/console but even this sometimes takes several minutes (if not dozens) just to show what’s going on.
It’s always a bit stressful to see the codes passing one after the other and then it stops on one and seems to get stuck there for a while before continuing the IPL process. Maybe it’s applying PTFs (updates) or something, and you just have to wait while even the console is blank.
I’ve been monitoring those servers for years and I’m still sometimes wondering if it hanged during the IPL or if it’s just doing its thing, because this part, even with codes, is not very verbose.
Fortunately it’s also very stable so it pretty much always comes back a few minutes after you start wondering why the hell it’s taking so long.
Dell PowerEdge R620, I’m talking to you.
When someone previously told a vrtx vm not to auto boot after power up and none of the remote access is working either… Both undocumented as well, of course. And your tired AF tech is statically configuring the wrong IP range on their laptop to manu because it’s been a long shutdown day and are also unfamiliar with the system in general (me). Good times, I figured it out though, but lots of sweating and swearing.
Ubuntu server just asked me if I want to upgrade to V24, I don’t know when I’ll take time to do that :p
In the old days some of the servers took at hour to reboot. That was stressful when you couldn’t ping it at an hour.
Don’t say stuff like that. You’re gonna give me a heart attack.
The more disk you had, the longer it took. It walked the scsi bus which took forever. So if you had more disk. It took even longer.
Since everything was remote, you’d have to call hands and they weren’t technical. Also no cameras since it was the 90’s.
Now when I restart a vm or container. I panic if it’s not back up in 10 minutes.
Why would you design a disk driver that way?
It isn’t a disk driver since the OS is not loaded yet. It is the hardware identifying each disk in the SCSI chain. Not sure what else it was doing walking the bus much I know finding all the disk was the longest part.
Shoot, it did occur to me that might not technically be the right word.
Still, even if you’re an engineer in the late 80’s, it seems like it would be obvious you need a way for disks to announce themselves in O(1) time. Was it just a limitation of interoperability between vendors or something?
I think it was just a limit of how quick everything ran back then. Also, this was an IBM system that was checked, double-checked, and triple-checked because it was a mission-critical system. IBM used to be known for quality hardware. Hard to imagine because they are such a crap company now but that was the equivalent of a google back then.
I get annoyed if my pc isn’t restarted in 30 seconds now.
I think mine takes like 2 minutes. It’s ten years old. I’ve putting off upgrading to the cost of videos cards
I got an M.2 drive last year after having a motherboard capable of it for 3-4 years, and naturally named it “Plash Speed”.
I will never not laugh at this video.
I like how posting got fairly fast. Then we started putting absurd amounts of ram into servers so now they’re back to slow.
Like we have a high clock speed dual 32 core AMD server with 1TB of ram that takes at least 5 minutes to do it’s RAM check. So every time you need to reboot you’re just sitting there twiddling your thumbs waiting anxiously.
I will date myself. These machines had a lot of memory as well which added to the slow reboot. I think it was 16 gigs.
The r series for IBM took forever. The p series was faster but was still slow
I’ll date myself. My first PC had 500MB of STORAGE
My first pc had a tape drive.
I had a friend with one of those while I had an Atari. The Atari game would come up within a minute, but the tape took like 15 min to start.
Using a tape drive is crazy when you think about it. It was slow…. This wasn’t the big tape cartridges. It was a standard Audio tape. Not sure why they could store but it was all sequential
Never ask an engineer why lol
Source: am engineer
this week i sudo shutdown now our main service right at the end of the workday because i tought it was a local terminal.
not a bright move.
Happens to everyone
Just having a multitude of terminals open with a mix of test environment and (just for comparison) an open connection to the production servers…
We were at a fair/exhibition once and on the first day people working on an actual customer project asked us, if they could compare with our code.
Obviously they flashed the wrong PLC and we were stuck dead at the first hours of the exhibition.
I still think that this place was cursed, as we also had to do multiple re-soldering of some connections of our robot and the sherry on top was the system flash dying - where I had fucked up, because I just finished everything late at night and didn’t made a complete backup of everything.
But it seems, if luck runs out, you lose on all fronts.At least I was able to restore everything in 20mins. Which must be some kind of record.
But I was shaking so much from the stress, that I couldn’t efficiently type anymore and was lucky to have a colleague to just calmly enter what I told him to and with that we’re able to get the show case up and running again.Well, at least the beer afterwards tasted like the liquid of the gods
I was making after hours config changes on a pair of mostly-but-not-entirely redundant Cisco L3 switches which basically controlled the entire network at that location. While updating the running configs I mixed up which ssh session was which switch and accidentally gave both switches the same IP address, and before I noticed the error I copied the running config to the startup config.
Due to other limitations and the fact that these changes were to fix DNS issues (and therefore I couldn’t rely on DNS to save me) I ended up keeping sshing in by IP until I got the right switch and trying to make the change before my session died due to dropped packets from the mucked up network situation I had created. That easily added a couple of hours of cleanup to the maintainence I was doing
Oops.
Since you’re using sudo, I suggest setting different passwords on production, remote, and personal systems. That way, you’ll get a password error before a tired/distracted command executes in the wrong terminal.
i have different passwords but i type them so naturally it didnt even register. so very dumb.
There’s a package called
molly-guardwhich will check to see if you are connected via ssh when you try to shut it down. If you are, it will ask you for the hostname of the system to make sure you’re shutting down the right one.Very usefull program to just throw onto servers.
nice. got it installed to test it out
We got the Trojan in, let’s move move move!
Best thing I did was change my shell prompt so I can easily tell when it isn’t my machine
you mean the user@machine:$ thing? how do you have yours?
Correct!
I put a little Home icon on mine using NerdFonts.
If you are using ZSH or Fish you can do much moreChange the color too
That’s why you connect an arduino to the motherboards reset pin and load it with a program where it resets the system if it doesn’t receive an ACK signal over the usb connection every 10 minutes.
Eventually though the networking and apache stops working after around 150 days so you also have to make a script that resets the system after 30 minutes of not having network.
Plot twist, reboot takes 11 minutes and you didn’t test for it
Cattle! Not pets!
