When I run index with the -N option I keep 
getting many "waiting for resolver" messages, and 
so have started using the -R option with the same 
number of resolvers as threads. (usually 15)

However, when running like this after about 10 
minutes everything seems to halt and I get 
an "actual gethostbyname timeout: xxx.xxx" error, 
where xxx.xxx appears to be half of a dotted quad 
(eg 157.234.) - I assume that this is infact 
related to a host the resolver is looking up, 
rather than thread/resolver/process id.

What does this error mean? and how can I stop it?

It usually shows one of these errors for each 
thread/resolver and then the index process 
appears to hang (although top still shows it 
using ~90% CPU time). Even running index -E in 
another shell has no effect, and I usually have 
to kill the process with a SIGKILL (often causing 
DB corruption).

Any ideas? if it helps, I'm running FreeBSD 4.3

-------------------------------------------------
This mail sent through UK Online webmail

Reply via email to