Join devRant
Do all the things like
++ or -- rants, post your own rants, comment on others' rants and build your customized dev avatar
Sign Up
Pipeless API
From the creators of devRant, Pipeless lets you power real-time personalized recommendations and activity feeds using a simple API
Learn More
-
@Oktokolo First, I blame me ! Why ? Because I already have a workaround. But this API skipped my radar for that.
Second, the load balacing is Azure Web App Plan Scale out feture. It's working great, except one thing : It balaces load based only on requests per second. And you can't tell it to use another metric.
I hope it is in complaince with guidelines now -
@NoToJavaScript
Is that true? Curious, cause that's really one of the worst design decision I've seen.
Requests per seconds is the most inaccurate way of load balancing or scheduling.
Requests in queue / active connections make more sense, as it is near to impossible that *every* request has the same duration time.
Which is why I absolutely hate when people start screaming requests per seconds with obscure numbers like it means something...
(It means someone is making dumb claims, yes - but otherwise it's hot air without content) -
@IntrusionCM I completly agree.
So, I didn’t look into it for couple of years and now there is a tiny tiny section in the doc for using “Health check”.
Relevant doc : https://learn.microsoft.com/en-us/...
But it’s only to avoid sending traffic to “crashed” or “Still booting” insurances, as the check is only done once a minute.
So as far as I can tell it’s still Round-robin.
And from what I see : You can put another load balkancer, but in this case you lose auto scaling capability.
Hey, I’m gonna create a support ticket. Let’s see what they respond. -
@IntrusionCM And to be pricise :
I'm talking about "Web plan scale out" feature.
They have multiple other offerings for load balacing and trafic management ! -
@IntrusionCM here youy go, confirmation from Azure support. You can't influance load balancer in any way
I'm hal laughting, half facepaling.
Use multiple servers they said
It's gonna balance automaticlly your load they said.
Well, they didn';t accoubnt for a MEGA badly writen API whic can block a whole server for a minute.
Or, but it's not the end !!!
So you see ? 1 at 100%, 1 at 0% ? Quick maths : Average 50%.
Autoscaling kicks in at average 40% for 2 minutes.
So it kicks in... Created a 3d server... Which is still booting !!
And some trafic redirected there.
Sooo. Users on a 100% server : Bad bad bad for them
Users on 0% server : No problems
users on a new 3d server : Wait 15 seconds, then it's like 0%.
Fun fun fun. At least I know now which fucking API does that.
rant