If you're talking about keep-alive this times out far before most of the DNS TTLs mentioned in this article and don't persist after a connection is closed anyways.
This is true in that the range encapsulates most keep-alive timeouts not in that keepalives longer than a minute or two are actually the majority. nginx defaults to 100 seconds, Apache is less than that. Most don't mess with these let alone bump them to 900. Generally 60 to 120 is considered standard with some cap on the number of active keep-alive sessions as well. Some go ultra-low or disable it all together, very few go ultra-high.
> WTF is your timeout?
120.
Also please try to keep the conversations together.
If you're talking about keep-alive this times out far before most of the DNS TTLs mentioned in this article and don't persist after a connection is closed anyways.
That's not true. It all depends on the implementation...
Also what are you talking about? If the connection is closed, how would it be used?
Connections should only be closed due to? Inactivity. If a connection is closed, don't you think you would probably want to do another DNS request?
Also if your doing proper layer 4 load balancing using BGP, DNS is a moot point...
Magic...
> Connections should only be closed due to? Inactivity.
Or if the user closes the browser of if the server/proxy restarts. But yes, mostly inactivity on the order of a couple of minutes.
> If a connection is closed, don't you think you would probably want to do another DNS request?
That's the whole point of the DNS TTL, to say how long to go before doing another lookup rather than doing it each time you reconnect.
> Also if your doing proper layer 4 load balancing using BGP, DNS is a moot point...
BGP load balancing operates on layer 3, and is irrelevant as you still need to DNS lookup an anycast address. EDNS client subnet is better anyways.
2 replies →
Also, wtf are you talking about? Did you even read the article?
Most sit between? 0-15 minutes. WTF is your timeout?
> Also, wtf are you talking about?
TTLS.
> Did you even read the article?
Sure did which is why I referenced it.
> Most sit between? 0-15 minutes.
This is true in that the range encapsulates most keep-alive timeouts not in that keepalives longer than a minute or two are actually the majority. nginx defaults to 100 seconds, Apache is less than that. Most don't mess with these let alone bump them to 900. Generally 60 to 120 is considered standard with some cap on the number of active keep-alive sessions as well. Some go ultra-low or disable it all together, very few go ultra-high.
> WTF is your timeout?
120.
Also please try to keep the conversations together.
2 replies →