Monty Ree wrote: > Hello, list. > > I have setup bayes related at local.cf like below. > > SA is 3.1.5 at linux. > > ## local.cf ## use_bayes 1 > bayes_path /data/.spamassassin/bayes > auto_whitelist_path /data/.spamassassin/auto-whitelist > bayes_file_mode 777 > bayes_auto_learn 1 > auto_whitelist_file_mode 777 > > Question. > -. At somt server, there are lots of bayes_toks.expire.xxxx files, > what is this? Temp files created during expiry... sounds like something keeps killing SA when it tries to expire your bayes DB thinking it's hung up and therefore "times out".
This is a common problem for MailScanner users, but it happens in any API layer tool with a "timeout" value for SA processing. SA by default tries to automatically do expiry at message processing time when needed, but this can cause that one message to take several minutes to process. Assuming no expiry is going on at the moment, these files can be deleted. > Curious thing is, some serves exist, but not at some servers. xxx is > random number. Actually it should be a process ID. > -. bayes_seen and bayes_toks are over 40M and still increases. Is > there any method or way to limit by some size, like 20M? Yeah, adjust the timeout of whatever is killing SA while it is trying to expire old tokens out of the bayes DB. In the short term, you can force-run an expire by running sa-learn --force-expire. > I 'm afraid SA can make some load because of this big size. > > > > Thanks for your time. > > _________________________________________________________________ > 인터넷 쇼핑은 MSN 에서 시작하세요. http://www.msn.co.kr/shopping >
