Steam outages every tuesday is dumb

[Technetheus]

What year is it anyway and why hasn't anything changed to mitigate the outages or remove them from a customer perspective. This is the worst part about steam.

[Lokey]

Messaggio originale di _galaxy:

Messaggio originale di Technetheus:

I'm against halfass system architecture and modern cloud services usually don't have weekly outages that impact customers.

Steam is in no way half assed also this isn't a "outage" it's a scheduled maintenace. It only sucks that it happens weekly.

An outage by definition is when power supply or other services are unavailable. Whenever a power plant goes offline for scheduled maintenance, it is an outage. Just because you don't agree with the defined term doesn't make it correct, nor do your quotation marks.

[Satoru]

Messaggio originale di The Commendatore:

Let's change it to Fridays.

Said no one who is in operations because you know 'nothing' ever goes wrong on a Friday night right?

[Technetheus]

Just saying often there are ways to mitigate this kind of problem. There's a difference between maintenance that is apparent to customers and maintenance that isn't. Embrace abstraction layers and decouple domains harder.

[_galaxy]

Messaggio originale di Technetheus:

Just saying often there are ways to mitigate this kind of problem. There's a difference between maintenance that is apparent to customers and maintenance that isn't. Embrace abstraction layers and decouple domains harder. :sunglassesDoge:

This lol.

But I'm gonna say... it's not a pretty solution, but IS a solution. I'm a software engineer, I know these kind of things are tough. It affects me on my free time too but it is what it is (at least for me).

[Zee_]

Sure there are ways they can do maintenance without taking everything offline, but heck, Steam is still a 32 bit application...

[nicole.macphee]

so lets ask the real question. when does it start working again for everyone?

[Sifer2]

Messaggio originale di rawWwRrr:

It's the one day and one time during that day that user loads are at their lowest. Seems the most optimal day and time to do them.

Whenever this comes up someone always says that but that window of time just happens to be prime time on the US East Coast. Right around the time many are home from work, and ready to game. Tuesday might be a slow day but it's hardly the slowest time if you live in the EST region. I lean more toward this just happens to be the time that's most convenient for Valve employees. They don't have to stay late for work or do it during early AM hours. Just a nice comfy little Tuesday afternoon maintenance after their lunch break.

[Leonardo Da Pinchi]

Messaggio originale di Technetheus:

I'm against halfass system architecture and modern cloud services usually don't have weekly outages that impact customers.

No, instead they tend to compile errors until they go down for days at a time.

[Satoru]

Messaggio originale di Technetheus:

Just saying often there are ways to mitigate this kind of problem. There's a difference between maintenance that is apparent to customers and maintenance that isn't. Embrace abstraction layers and decouple domains harder. :sunglassesDoge:

I mean yes but also no. While you might be able to do rolling reboots for certain things, lots of stuff simply cannot be done that way. Also you have the issue where you cannot maintain sessions when you reboot a system a user is on. Yes some things are stateless, but a lot of stuff isn't.

You also have the issue of trying to not only send new connections to only 'upgraded' servers, which will, but also migrate existing sessions to them. This will, pretty much inevitably, overload any upgraded servers within the first few servers you do causing a massive cascade effect where by you basically have to kick everyone off all the other non-upgraded servers anyway because you simply don't have the capacity anymore. Not everyone has 2x capacity just lying around for outages as this is extremely expensive.

[_galaxy]

Messaggio originale di Satoru:

Messaggio originale di Technetheus:

Just saying often there are ways to mitigate this kind of problem. There's a difference between maintenance that is apparent to customers and maintenance that isn't. Embrace abstraction layers and decouple domains harder. :sunglassesDoge:

I mean yes but also no. While you might be able to do rolling reboots for certain things, lots of stuff simply cannot be done that way. Also you have the issue where you cannot maintain sessions when you reboot a system a user is on. Yes some things are stateless, but a lot of stuff isn't.

You also have the issue of trying to not only send new connections to only 'upgraded' servers, which will, but also migrate existing sessions to them. This will, pretty much inevitably, overload any upgraded servers within the first few servers you do causing a massive cascade effect where by you basically have to kick everyone off all the other non-upgraded servers anyway because you simply don't have the capacity anymore. Not everyone has 2x capacity just lying around for outages as this is extremely expensive.

Steam has the resources to fix this problem if they wanted (which means figuring out a good solution as well), they deem the current solution good enough, it will come a point however where steam is so big the outage money loss would be substantial enough that they will have to deal with it