[Mimedefang] Re: R/W: lock failed: File exists

[Lee Dilkie]

>> It is trying to lock the bayes journal, and other mimedefang 
>threads can't
>> read to it while it learns from the journal.
>> This will solve your problem.
>> 
>> My settings:
>> bayes settings:
>> bayes_learn_to_journal 1
>> bayes_journal_max_size 5120000
>> 
>> 
>> cronjob:
>> */55 * * * *        su -c 'sa-learn --rebuild' defang; echo defang >
>> /dev/null
>
>Very well, but...
>
>What you suggest is a kinda workaround to be applied at the 
>sites where the
>problem occurs. Now, since the nature of the problem seems to be quite
>generic I'm just very curious why don't *others* complain??! Because it
>looks as if only a *minority* of MD/SA installations (like 
>yours and mine)
>get those nasty messages and so you offer a workaround for them.
>
>So now, just for my curiosity sake, I have a very basic 
>question: which of
>these two points below is true and which is false.
>
>1. There is something weird with my installation but, OK, luckily 
>   enough there's a workaround.
>2. Everyone has the same problem so for each and every MD/SA 
>installation
>   a workaround like that is needed.
>   
>And finally, is 2. is true then maybe something more than a workaround 
>should be invented?

Well, I've had this problem since forever. It started a month or so after I turned Bayes on. Like you I did an investigation and my conclusion what that it wasn't particularly harmful... Although I don't think I arrived at the same conclusion as you folks. I seem to recall that the problem was that one bayes process (MD) had the file locked when the other bayes process (MD) was trying to trim out stale entries from the db and wasn't able to since the file was locked. My reading on the matter suggested it wasn't a big deal as the trim would occur again later. I didn't read anything that would suggest it actually interfered with the operation of bayes in the first place.. Perhaps I was wrong in that?

-lee