[dnsdist] setAuto() healthchecks

[Remi Gacogne]

On 08/17/2017 05:15 PM, Thib D wrote:
>     This looks wrong to me, and I just opened a PR [1] that sets the backend
>     state to down when a setAuto() is issued. It means the backend will stay
>     down until the next heath check, as it does when it is first added.
>     Let me know if it solves your issue or if I didn't understand it
>     correctly.
> 
> 
> In this case, this would be solving our issue. 
> 
> But does it means you would be considering a server as "down"  everytime
> you use setAuto() ? What if the server is actually up ?

Until the next health check, yes, so for roughly one second. Note that's
also what we do when the server is first added, its initial state is
"down" until its health has been checked (unless it's forced  up via
setUp() of course).
I agree this can be an issue if the server was forced up and you want to
enable back health checks, because you then have a one second downtime.
Perhaps we should add an optional parameter to setAuto() to be able to
set the initial state. No parameter would be the current behavior (ie,
previous state), "false" would mean the server is marked down until the
next check and "true" marked up instead.

-- 
Remi Gacogne
PowerDNS.COM BV - https://www.powerdns.com/

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://mailman.powerdns.com/pipermail/dnsdist/attachments/20170817/3a45d361/attachment.sig>