Bind broken after CentOS 5.4 update

[smoge]

Broken system(s)

Name : bind-libs
Arch : i386
Epoch : 30
Version : 9.3.6
Release : 4.P1.el5

Name : bind-libs
Arch : x86_64
Epoch : 30
Version : 9.3.6
Release : 4.P1.el5

Working system(s)

Name : bind-libs
Arch : i386
Epoch : 30
Version : 9.3.4
Release : 10.P1.el5_3.3

Name : bind-libs
Arch : x86_64
Epoch : 30
Version : 9.3.4
Release : 10.P1.el5_3.3

Name : bind
Arch : x86_64
Epoch : 30
Version : 9.3.4
Release : 10.P1.el5_3.3

[smoge]

Doing stuff mentioned before hand in this ticket - does not help... so trying some things myself....

I turned off AUTO UPDATES on all our other servers as I assume this was something that came through in recent updates...

[smoge]

Softlayer ( Welcome to SoftLayer ) - which is an excellent hosting company, by the way, and I , figured out a temp fix for this... I have a ticket open with CPANEL to see if they approve of it - but it does seem to work...

We seem to have fixed this - by adding in /usr/lib64

the following symlinks

ln -s libisc.so libisc.so.11
ln -s libdns.so libdns.so.22

Smoge

[Varial]

Bind appears to be broken after tonight's automatic upgrade to CentOS 5.4. Any quick fixes?

[Varial]

I wasn't able to get this working by symlinking like you have suggested. It appears the new version of Bind in CentOS 5.4 breaks things.

[Solokron]

We are seeing the same thing. I have tried the sym links as well with a service named restart to no avail.

[jagfly]

Same Issue here. Bind died at 4AM after an update and has been unable to restart despite trying all the usual fixes. Eagerly awaiting a resolution.

[Solokron]

/scripts/fixndc
Named could not be restarted, any obvious config errors should show up below this line.
No critical problems found, will attempt to regenerate keys regardless.
warn [fixrndc] /usr/sbin/rndc status failed: rndc: connect failed: 127.0.0.1#953: connection refused
Creating rndc.conf
Creating /etc/rndc.key
warn [fixrndc] /usr/sbin/rndc status failed: rndc: connect failed: 127.0.0.1#953: connection refused
Restarting named
warn [fixrndc] /usr/sbin/rndc status failed: rndc: connect failed: 127.0.0.1#953: connection refused
/scripts/fixrndc failed to fix the rndc key (or named is otherwise broken), please investigate manually

[smoge]

We seem to have fixed this - by adding in /usr/lib64

the following symlinks

ln -s libisc.so libisc.so.11
ln -s libdns.so libdns.so.22

Smoge

[smoge]

We just had another server go down... and the

We seem to have fixed this - by adding in /usr/lib64

the following symlinks

ln -s libisc.so libisc.so.11
ln -s libdns.so libdns.so.22

and a

service named restart

did the trick.

[Solokron]

cd /usr/lib64
ln -s libisc.so libisc.so.11
ln -s libdns.so libdns.so.22

service named restart

if still broken backup /etc/named.conf and disable bind in WHM, save and then re-enable.

[darren.nolan]

Same issue with only 1 of my cPanel servers.

Support Request ID 510973.

[darren.nolan]

cPanel (being awesome as always) tracked down my particular issue to invalid zone files.

/etc/init.d/named start | grep -v "loaded serial"

On the previous version of Bind, these little mistakes in the zone files were ignored, latest version - not so much. Zone files updated/fixed and everything works fine.

This would explain why only one out of several of my servers were affected by the update.

[cPanelDon]

Darren is correct; the new BIND update ensures more strict RFC compliance, where BIND configuration and DNS zone entries must have proper syntax.

The command shown by Darren was used to help filter some of the output during BIND/named startup so that the errors could be discerned and investigated with more precision and efficiency.

# /etc/init.d/named start | grep -v "loaded serial"

Note: I have merged posts from another topic into this thread where they can receive more relevant discussion.

If anyone else is experiencing difficulty tracking down why BIND is not restarting please consider submitting a ticket; we are here 24/7 and will be able to assist with corrective measures where applicable. You may submit a ticket via the link in my forums signature.

[VeZoZ]

What is causing it to check more is in /etc/init.d/named there is this line:

ckcf_options='-z'; # enable named-checkzone for each zone (9.3.1+) !

So you could fix the zones or alternatively comment out that line so the broken zones do not prevent bind from starting.