You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by Matias Lopez Bergero <ml...@udesa.edu.ar> on 2005/02/08 20:37:50 UTC
bayes db version error
Hi
I'm seeing a lot of messages about and version error in the bayes db in
my log file:
spamd[6562]: bayes: bayes db version 0 is not able to be used, aborting!
at /usr/lib/perl5/site_perl/5.8.0/Mail/SpamAssassin/BayesStore/DBM.pm
line 160.
I'm using SA 3.0.2.
The output of sa-learn show a bayes db version 3.
sa-learn --dump magic | grep version
0.000 0 3 0 non-token data: bayes db version
This is correct? The reported error it's about version 0.
I have read trough the list that this problem is caused by the update of
the spamassassin version.
I start with the SA version 2.55 that was carried default with the
distro. Then upgrade to 3.0.1 or 3.0.0, so 3.0.1 and now 3.0.2. Maybe
soon 3.0.3? ;)
There is a way to correct this problem? I was thinking in delete the old
db, but I'm not very sure where to look. I not use the 2.55 version of
SA, it was installed with the distro and I upgrade immediately after
install. I looked at /root/.spamassassin for old files but found
nothing. Where else this old db could be??
Could this be affecting the spam filtering?
BR,
Matías.
Re: bayes db version error
Posted by Matias Lopez Bergero <ml...@udesa.edu.ar>.
Michael Parker wrote:
> On Tue, Feb 08, 2005 at 05:28:51PM -0300, Matias Lopez Bergero wrote:
>
>>>2) Throttle the calls to spamd to reduce lock contention.
>>
>>Sorry to ask this again, but I'm not native English speaker :-P
>>Did you mean to increase the number of spamd children processes, like
>>spamd -m x? I have currently set spamd -m 10.
>>
>
> No, it means to slow/reduce the calls to spamd. Increasing the number
> of children will probably make the problem get worse.
OK, I'm using milter-spamc to talk with sendmail milter and pass the
messages to spamd. For what I know, there is no way to control the calls
to spamd from the milter-spamc command.
I would have to reduce the spamd child processes or increase the milter
timeout in order to reduce the calls to spamd right?
>>>3) Switch to SQL based bayes which won't (well shouldn't) have that
>>> issue.
>>
>>That's an interesting idea.
>>I'm going to keep that in mind :)
>
> You can view the notes/slides from my ApacheCon presentation on
> Storing SpamAssassin User Data in SQL Databases here:
> http://www.apache.org/~parker/presentations/
>
> Hopefully it will help move things along.
That's good,
Thank you very much Michael.
BR,
Matías.
Re: bayes db version error
Posted by Michael Parker <pa...@pobox.com>.
On Tue, Feb 08, 2005 at 05:28:51PM -0300, Matias Lopez Bergero wrote:
>
> >2) Throttle the calls to spamd to reduce lock contention.
>
> Sorry to ask this again, but I'm not native English speaker :-P
> Did you mean to increase the number of spamd children processes, like
> spamd -m x? I have currently set spamd -m 10.
>
No, it means to slow/reduce the calls to spamd. Increasing the number
of children will probably make the problem get worse.
>
> >3) Switch to SQL based bayes which won't (well shouldn't) have that
> > issue.
>
> That's an interesting idea.
> I'm going to keep that in mind :)
You can view the notes/slides from my ApacheCon presentation on
Storing SpamAssassin User Data in SQL Databases here:
http://www.apache.org/~parker/presentations/
Hopefully it will help move things along.
Michael
Re: bayes db version error
Posted by Matias Lopez Bergero <ml...@udesa.edu.ar>.
Michael Parker wrote:
> On Tue, Feb 08, 2005 at 04:37:50PM -0300, Matias Lopez Bergero wrote:
>
>>I'm seeing a lot of messages about and version error in the bayes db in
>>my log file:
>>
>>spamd[6562]: bayes: bayes db version 0 is not able to be used, aborting!
>>at /usr/lib/perl5/site_perl/5.8.0/Mail/SpamAssassin/BayesStore/DBM.pm
>>line 160.
>>
>
>
> I'm assuming you're running sitewide bayes (or at least running as a
> single user) and on a somewhat busy server.
Yes, I forgot to say that. I'm running a sitewide install with about
60000 incoming messages per day.
> That error message is a pretty good indication that SA couldn't get a
> lock on the bayes db files. It's actually just a warning, not an
> error, and it may or may not have actually aborted.
>
> You'll see this on a setup that is getting a good amount of traffic
> and using shared/sitewide bayes db files.
>
> Several things you can try:
>
> 1) If you db files aren't on an NFS filesystem switch your lock_method
> to flock (the default is nfssafe). If your shared db files are on
> an NFS filesystem then consider moving them off and switch your
> lock_method.
Done.
> 2) Throttle the calls to spamd to reduce lock contention.
Sorry to ask this again, but I'm not native English speaker :-P
Did you mean to increase the number of spamd children processes, like
spamd -m x? I have currently set spamd -m 10.
> 3) Switch to SQL based bayes which won't (well shouldn't) have that
> issue.
That's an interesting idea.
I'm going to keep that in mind :)
Thanks a lot Michael
BR,
Matías.
Re: bayes db version error
Posted by Michael Parker <pa...@pobox.com>.
On Tue, Feb 08, 2005 at 04:37:50PM -0300, Matias Lopez Bergero wrote:
> I'm seeing a lot of messages about and version error in the bayes db in
> my log file:
>
> spamd[6562]: bayes: bayes db version 0 is not able to be used, aborting!
> at /usr/lib/perl5/site_perl/5.8.0/Mail/SpamAssassin/BayesStore/DBM.pm
> line 160.
>
I'm assuming you're running sitewide bayes (or at least running as a
single user) and on a somewhat busy server.
That error message is a pretty good indication that SA couldn't get a
lock on the bayes db files. It's actually just a warning, not an
error, and it may or may not have actually aborted.
You'll see this on a setup that is getting a good amount of traffic
and using shared/sitewide bayes db files.
Several things you can try:
1) If you db files aren't on an NFS filesystem switch your lock_method
to flock (the default is nfssafe). If your shared db files are on
an NFS filesystem then consider moving them off and switch your
lock_method.
2) Throttle the calls to spamd to reduce lock contention.
3) Switch to SQL based bayes which won't (well shouldn't) have that
issue.
>
> Could this be affecting the spam filtering?
>
In theory, things should filter just fine, you just won't get BAYES
results. If you're seeing something different then it's probably a
bug.
Michael