What do servers do when they go down for maintenance? Do they defrag the hard drives or something?

What do servers do when they go down for maintenance? Do they defrag the hard drives or something?

Attached: ryzen.jpg (1200x953, 214K)

They get a free meal from their bosss

They don't go down for "maintenance", they go down cause something breaks and needs to be replaced, like ram.

Firmware updates most of the times. Either to fix bugs or to patch vulnerabilities.

They defrag the RAM and overclock the BMC.

In Linux it's due to needing kernel patches or hardware/firmware issues. In Windows there are several reasons including "because it shit itself".

You can patch the kernel without rebooting the system now.

I think he is asking about game servers like how wow servers go down once a week.

honestly its prob due to some memory leak or flaw in blizzard libtard code.

probably updating windows

Attached: wu_top-650x300.png (650x300, 11K)

this is a good question that I've wondered myself..
why is this even a thing?

the rest of the IT world does maintenance, patching, deployments etc without downtime, because downtime is unacceptable.

why is the gaming industry the only place where server downtime for HOURS is totally normal?

I imagine with the load they're working with eveything is loadbalanced and distributed over multiple servers anyway, so why is it they can't gradually update/patch/upgrade/deploy stuff node per node ? or use immutable deploys?

don't tell me they run everything on physical machines that're manually managed...

Preloading new game content, kick off everyone so they are forced to update and sync versions before login on again.

most game server farms look like this

Attached: wrong4.jpg (640x480, 63K)

It is mostly so you can backup data. Also lots of game server's software require binary fixes, and you "can't" really do that in memory/realtime, so it requires you to bring the servers down to apply them. In the case of pushing out client patches, it also lets you refresh the playerbase and force them to update.

still, that should only take like half an hour at most, you can do all that beforehand and just switch over to the new stuff in a matter of minutes.

I wonder that too. How is blue-green deployments not a thing?

Not necessarily, sometimes you have hundreds of GB of databases you need to back up, which can take a significant amount of time. Also, not many gaming companies use a staging environment (some do, just not all) they just apply the "patch" directly onto the production environment. On occasion too, if even when using a staging environment setup to be a direct copy of production, issues can pop up that you didn't forsee requiring you to fix or rollback the changes which of course takes extra time.
t. guy who does the game server maintenance.

>Not turning them off because of a "problem" once in a while to remind management they actually need you and not outsource to pajeets

Blue-green does not work for realtime streaming connections. Back to school, user

all the good devs work for payment processing where these issues magically do not affect consumers

Not the full thing per se. But rather, why don't u just keep your some green servers while u await to spin up the blue ones to certain amounts before killing all the green ones? Since the old version server is killed off, clients would disconnect. Thereafter, reroute all traffic to the newer version and clients are forced to update before connecting.

This pattern is not uncommon. Some games like Team Fortress 2 does it (in a way). Valve releases the update, both client and server has to update. If server and client isn't same version, then they can't connect.

At the very least they should do it for minor patches where the state of the characters just need to be "casually consistent". If there are major patches, I think it would be understandable to maintain strongly consistent states.

You can not be serious if you believe payment processors have a 100% uptime, and never run into issues during a maintenance. It is also a fundamentally different infrastructure than commercial gaming server systems.

Lazy admins just reboot the machine and keep it down for a while to pretend they are doing a job in order to reduce the amount of labor they perform while simultaneously maintaining (or even increasing) their income.

They're pwning us all and we don't even realize.

Attached: flutterwlan0_by_rapidstrike-daqx994.png (253x250, 52K)