named memory consumption problem

[Dmitry M.]

Hi,

I have 3 DNS servers, let say ns1 ns2 ns3, all of them uses DNSonly cpanel with standard bind 9.3.*

ns2 and ns3 are fully identical 32 bit CentOS release 5.2
ns1 is 64bit CentOS release 5.4

ns2 and ns3 works for 10 months without any problem
ns1 is new server.

The problem is named on ns1 uses 80% more RAM than named on ns2 or ns3.

I copied named.conf from ns2, so the config isn't the problem. DNS zones were also copied from dns2.

Any ideas how to fix the RAM consumption on ns1?

P.S. I also have the 4th DNS server (old ns1), this server uses exact the same amount RAM that ns2/ns3 uses, but it runs under old CentOS 4.x and this server cannot be upgraded with more RAM (old hardware).

[DjiXas]

Hi,

I have 3 DNS servers, let say ns1 ns2 ns3, all of them uses DNSonly cpanel with standard bind 9.3.*

ns2 and ns3 are fully identical 32 bit CentOS release 5.2
ns1 is 64bit CentOS release 5.4

ns2 and ns3 works for 10 months without any problem
ns1 is new server.

The problem is named on ns1 uses 80% more RAM than named on ns2 or ns3.

I copied named.conf from ns2, so the config isn't the problem. DNS zones were also copied from dns2.

Any ideas how to fix the RAM consumption on ns1?

P.S. I also have the 4th DNS server (old ns1), this server uses exact the same amount RAM that ns2/ns3 uses, but it runs under old CentOS 4.x and this server cannot be upgraded with more RAM (old hardware).

I remember having similar issue (just with CPU) when I accidentally added ; into one of the records.

[Dmitry M.]

I remember having similar issue (just with CPU) when I accidentally added ; into one of the records.

I have already tried
1. Fully remove all DNS zones and copy new ones from dns2/dns3 (3 times)
2. I checked /var/log/message and /var/named/data/named.run for bad zones (found and removed one zone with full re-sync of all DNS zones)

80% more is a huge amount of memory, don't think it is caused by some 1 zone. My named uses more than 1Gb of RAM, as you understand 1 or even 10-20 bad zones cannot cause 800Mb "named" size increase. I think.