Derek Catanzaro wrote:
>
>>  
> I am running sa-learn --force-expire -D on one of my servers now.  I
> mentioned above the the bayes_toks file was 321 MB.  Well on the
> server I am working on now it is 603 MB (unbelievable).  Anyway here
> is the out put from the -D.  Would I be better off deleting bayes_toks
> and starting over?  Or should I just let the expire run?
Your choice. sa-learn --clear will wipe the database in a lock-safe manner
> I did pull up another terminal from a different machine and I do not
> see a .expire file yet?  Does this file need to be in place already or
> is it created when you run sa-learn --force-expire?  
No, it will be created.. From the message below, sa-learn is still 
trying to figure out a good "cut-off date" for tokens.
> period: 43200
> [3990] dbg: bayes: can't use estimation method for expiry, unexpected
> result, calculating optimal atime delta (first pass)
> [3990] dbg: bayes: expiry max exponent: 9 

> OK, It finally started giving me more output after about 45 minutes.
> Now I am seeing the following lines repeating over and over again in
> the debug info.  The bayes_toks.expire file created and it is slowly
> growing  (at 20 MB now).  Do I need to be concerned with the repeating
> info below or is everything working as it should since the .expire
> file is growing?
>
> [3990] dbg: locker: refresh_lock: refresh /root/.spamassassin/bayes.mutex
> [3990] dbg: locker: refresh_lock: refresh /root/.spamassassin/bayes.mutex 
No, that's normal. sa-learn needs to refresh it's lock periodically or
other SA instances will just assume it crashed and take over the lock.
I'd be concerned if these messages stop coming.



Reply via email to