[Opendnssec-user] Problems adding largish # of zones

[Berry A.W. van Halderen]

On 12/03/2015 11:04 PM, Havard Eidnes wrote:
> In a somewhat desperate attempt to gather more information and
> remove some of the internal stumbling-blocks in OpenDNSSEC, I
> took a hatchet to the code.  I started around that obnoxious
> "transfer in progress" message and first created this patch:
[..]
> Add more debugging code to zero in on "transfer already in progress" errors.
> Also, don't skip sending notify messages to xfrd.
[..]
> The connections rise up to the limit (50):
> 
[..]
> connections are then released down to 0:
> 
[..]
> 
> Dec  3 21:42:26 ods-host ods-signerd: [xfrd] zone 2.39.128.in-addr.arpa skips retry: tcp connections full
> 
> What?!??  Who overwrote xfrd->tcp_waiting?!?
> 
> That field is only supposed to be set when we reach 50 active TCP
> sessions, and is cleared on every release action!

Dear Havard,

Great work to track down the problems in this area, this gives us the
hints we need to take this issue up and reproduce the stuck zones on
XFRs.  I can understand your state of mind because this indeed is not
easy to track down and takes up a lot of time to do so.

We're next try to reproduce the issue and also work on getting this
back on track.  It shows indeed a problem with this limit counter that
it isn't updated correctly.  We'll update on this.

With kind regards,
Berry van Halderen