outgoing AXFR fails with large zones

[anon]

We have a dns setup with ns1 served by powerdns, ns2 and ns3s are bind9, which are our slaves and get their zones by AXFR. PowerDNS is configured with the gmysql backend against a locally running mysql (4.1.11a-4sarge2).

With pdns r923 (also with r904) the AXFR fails for "large" zones (~550 records), but works well with smaller ones.

dig @ns1... AXFR large.zone shows after 301 records:

;; communications error: end of file

PDNS log shows:

Nov 21 15:08:24 ns1 pdns[18936]: AXFR of domain 'large.zone' initiated by cli.ent.ip.xxx
Nov 21 15:08:24 ns1 pdns[18936]: TCP server is without backend connections in doAXFR, launching
Nov 21 15:08:24 ns1 pdns[18936]: gmysql Connection succesful
Nov 21 15:08:24 ns1 pdns[18936]: gmysql Connection succesful
Nov 21 15:08:34 ns1 pdns[18936]: TCP nameserver had error, cycling backend: Trying to read data from remote TCP client cli.ent.ip.xxx: Resource temporarily unavailable

With 2.9.20 the AXFR worked, but after some time we saw that problem with tcp connections hanging open...

[anon]

fixes the mentioned axfr problems for me

[ahu]

Should be fixed by the recent commits (982 and preceeding)