[yadifa-users] yadifa 2.2.0 aggressive slave behaviour

yadifa info at yadifa.eu
Wed Jul 20 14:15:07 CEST 2016


Hello Anand,

> I don't understand something, so could you please clarify?

We apologize our previous explanation was insufficient. Additional information is coming in the documentation, but we hope this reply will make the behavior clear.

> Has that bug actually been fixed? A 600s delay between failed AXFR retries would already be a good thing, but it seems to be that it wasn't even doing that.

Yes, this issue has been fixed. The delay was not properly taken into account.

> Secondly, if I set the multiplier to 2, will that result in retry intervals values of 600s, 1200s, 2400s, 4800s, 9600s, ...? If so, then this isn't a linear fashion, it's exponential.

The formula is: axfr-retry-delay + axfr-retry-jitter + MIN(failure-count * axfr-retry-failure-delay-multiplier, axfr-retry-failure-delay-max)

When the multiplier is set to 2 (and ignoring the jitter), the retry intervals will be the values 602s (600s + 1*2), 604s (600 + 2*2),  606s (600 + 3*2),  ...
The increase is linear and not exponential. Currently the default value for axfr-retry-failure-delay-multiplier is set to 5 and the maximum is set to 3600. The default value for axfr-retry-failure-delay-max is set to 86400 and the maximum is set to 604800. We may tweak these values in the future.

> But then, the option isn't a multiplier, but an addend

If you have name which better fits the behavior of the feature, we will happily add it as an alias.

Kind regards,

R&D team
EURid VZW

-----Original Message-----
From: yadifa-users [mailto:yadifa-users-bounces at mailinglists.yadifa.eu] On Behalf Of Anand Buddhdev
Sent: Wednesday, July 20, 2016 12:03 PM
To: yadifa-users at mailinglists.yadifa.eu
Cc: anandb at ripe.net
Subject: Re: [yadifa-users] yadifa 2.2.0 aggressive slave behaviour

On 19/07/16 18:19, yadifa wrote:

Hi yadifa devs,

>> Could you please consider adding code to yadifa to do ... back-off
> 
> In release 2.2.1, we have added some back-off mechanism with configurable parameters:
> 
> axfr-retry-failure-delay-multiplier which will increase the back-off 
> time in a linear fashion axfr-retry-failure-delay-max which can be used to set the maximum time between failed retries.
> 
> More information about these parameters will be documented in the 2.2.1 reference manual (coming soon).

I don't understand something, so could you please clarify?

There is already an option, called "axfr-retry-delay" in yadifa, and it defaults to 600s. To me, this means that when AXFR for a zone fails, yadifa should wait 600s before trying again. However, when I tested 2.2.0, it wasn't even waiting 600s - it was retrying the failed zones immediately. Has that bug actually been fixed? A 600s delay between failed AXFR retries would already be a good thing, but it seems to be that it wasn't even doing that.

Secondly, if I set the multiplier to 2, will that result in retry intervals values of 600s, 1200s, 2400s, 4800s, 9600s, ...? If so, then this isn't a linear fashion, it's exponential. If you really meant linear, then you want to *add* to the retry each time, and have 600, 1200, 1800, 2400, 3200, etc. But then, the option isn't a multiplier, but an addend.

Therefore, please clarify what the actual behaviour is, and whether the option names matches that behaviour.

Regards,
Anand
_______________________________________________
yadifa-users mailing list
yadifa-users at mailinglists.yadifa.eu
http://www.yadifa.eu/mailman/listinfo/yadifa-users


More information about the yadifa-users mailing list