Re: question about potential feature enhancement

Dampening? 

delay starts at 1m (60)

// pseudo

int delay = 60;

..

..

int result = connect(XXXX);

if(result == -1) {

  delay = (delay < 300 ? delay+30:300);

  

} else {

   delay = 60;

}

---

From: "Saku Ytti via users" <users@conserver.com>
To: users@conserver.com
Sent: Friday, March 18, 2016 8:30:31 AM
Subject: question about potential feature enhancement

First, many thanks for conserver, I really love how it allows us to
use arbitrary console servers and yet get universal single UI access
to them.


Having said that, couple features I'd love to see, can someone be
convinced to code them, and if so, would upstream even accept them?

a) match-anywhere for console port name, instead of matching from
start of the string. Use case: your pop name is in the middle of the
string, and you want to check what consoles are available in the pop.

b) exponential autoreinit backoff. If there are lot of failing nodes,
1min may be woefully ineffective timeout, it would be nice if per
device delay would increase exponentially on subsequent failures.
Something like vendors implement for interface dampening (e.g.
http://www.cisco.com/c/en/us/td/docs/ios/12_0s/feature/guide/s_ipevdp.html)
this could completely replace retries, so instead of having n retries
+ reinit timeout, you'd just have single exponentially increasing
retry, eventually converging to to stable state even on busiest
system.
(Or simpler solution, configurable autoreinit delay)


Being C illiterate, I still might be able to produce patch for a) and
configurable autoreinit delay, but I can't imagine being able to
refactor the retry+autoreinit in single exponential backoff and have a
patch which would be accepted.

--
  ++ytti
_______________________________________________
users mailing list
users@conserver.com
https://www.conserver.com/mailman/listinfo/users