You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by snowcrash+sa <sc...@gmail.com> on 2007/09/19 17:34:47 UTC
(old?) sa hang @ "bayes: can't use estimation method for expiry"
Hi,
I'm updating SA & exploring/cleaning Bayes DBs on a slighty 'dusty' box.
after update to latest SA 32x-branch, --lint is OK.
but, currently, i get:
sa-learn --force-expire -D
...
[19416] dbg: bayes: found bayes db version 3
[19416] dbg: bayes: opportunistic call attempt skipped, found fresh
running expire magic token
[19416] dbg: config: score set 3 chosen.
[19416] dbg: learn: initializing learner
[19416] dbg: bayes: bayes journal sync starting
[19416] dbg: bayes: bayes journal sync completed
[19416] dbg: bayes: expiry starting
[19416] dbg: locker: safe_lock: created /etc/sa/local/.spamassassin/bayes.mutex
[19416] dbg: locker: safe_lock: trying to get lock on
/etc/sa/local/.spamassassin/bayes with 300 timeout
[19416] dbg: locker: safe_lock: link to
/etc/sa/local/.spamassassin/bayes.mutex: link ok
[19416] dbg: bayes: tie-ing to DB file R/W
/etc/sa/local/.spamassassin/bayes_toks
[19416] dbg: bayes: tie-ing to DB file R/W
/etc/sa/local/.spamassassin/bayes_seen
[19416] dbg: bayes: found bayes db version 3
[19416] dbg: locker: refresh_lock: refresh
/etc/sa/local/.spamassassin/bayes.mutex
[19416] dbg: bayes: expiry check keep size, 0.75 * max: 112500
[19416] dbg: bayes: token count: 218650, final goal reduction size: 106150
[19416] dbg: bayes: first pass? current: 1190215099, Last:
1110223381, atime: 1261902, count: 28215, newdelta: 335417, ratio:
3.76218323586745, period: 43200
[19416] dbg: bayes: can't use estimation method for expiry,
unexpected result, calculating optimal atime delta (first pass)
[19416] dbg: bayes: expiry max exponent: 9
where it just 'sits' for awhile -- no errors in logs -- and,
eventually, 'releases' and completes without error.
this only seems to happen an unquantified "some of the time" ...
sometimes (as moments ago) repeating the process results in no such
hand/delay.
fyi,
grep -i bayes local.cf
use_bayes 1
use_bayes_rules 1
bayes_auto_learn 1
bayes_min_ham_num 200
bayes_min_spam_num 200
bayes_learn_during_report 1
bayes_use_hapaxes 1
bayes_journal_max_size 102400 # in bytes
bayes_expiry_max_db_size 150000 # of tokens
bayes_auto_expire 1
bayes_learn_to_journal 1
bayes_path /etc/sa/local/.spamassassin/bayes
bayes_auto_learn_threshold_nonspam 1.0
bayes_auto_learn_threshold_spam 6.0
and,
cd /etc/sa/.spamassassin
ls -al
total 36024
drwxrwxr-x 10 sa sa 340 2007-09-19 08:22 .
drwxrwxr-x 39 sa sa 1326 2007-09-19 06:59 ..
-rw-rw---- 1 sa sa 90112 2007-09-19 08:23 auto-whitelist
-rw-rw---- 1 sa sa 10244096 2007-06-28 07:09 auto-whitelist-old
-rw-rw---- 1 sa sa 6 2007-09-19 08:23 auto-whitelist.mutex
-rw-rw---- 1 sa sa 30 2007-09-19 08:21 bayes.mutex
-rw------- 1 root sa 816 2007-09-19 08:22 bayes_journal
-rw-rw---- 1 sa sa 20570112 2007-09-19 07:30 bayes_seen
-rw-rw---- 1 sa sa 5488640 2007-09-19 08:21 bayes_toks
i had NOT noticed this before on this box ... not clear if it's the
build or my local.cf's bayes-related settings. or, perms, maybe?
i've found some old thread references to
"[19416] dbg: bayes: can't use estimation method for expiry, unexpected
result, calculating optimal atime delta (first pass)"
but not yet a resolution ... (still loooking!)
hints?
thanks.
Re: (old?) sa hang @ "bayes: can't use estimation method for expiry"
Posted by snowcrash+sa <sc...@gmail.com>.
> Yeah, that's the expiry first pass. As it said, it couldn't find a
> good atime delta to use for expiry, so it didn't anything.
ok, so it's "expected". read on ...
> "man sa-learn" has a large amount of information about how all of this works.
Arguably not in normal-human-english, but, yes.
The 'trigger' for my question is the presence of "unexpected result"
in the report.
No, it doesn't SAY "error" -- but neither does the manpage literallly
address "unexpected". It implies something outside of the norm. I'd
suggest something a bit clearer, especially given the existing 'old'
list-references to this *AS* an error, with nothing newer available
...
My $0.02. Thx.
Re: (old?) sa hang @ "bayes: can't use estimation method for expiry"
Posted by Theo Van Dinter <fe...@apache.org>.
On Wed, Sep 19, 2007 at 04:41:46PM -0700, snowcrash+sa wrote:
> > > where it just 'sits' for awhile -- no errors in logs -- and,
> > > eventually, 'releases' and completes without error.
> >
> > What are the messages when that happens? It shouldn't just exit.
>
> [19725] dbg: bayes: can't use estimation method for expiry,
> unexpected result, calculating optimal atime delta (first pass)
> (waits here ~ 2 minutes ...)
> [19725] dbg: bayes: expiry max exponent: 9
> [19725] dbg: bayes: atime token reduction
> [19725] dbg: bayes: ======== ===============
> [19725] dbg: bayes: 43200 221166
[...]
> [19725] dbg: bayes: 22118400 203758
> [19725] dbg: bayes: couldn't find a good delta atime, need more
> token difference, skipping expire
> [19725] dbg: bayes: expiry completed
> [19725] dbg: bayes: untie-ing
> [19725] dbg: bayes: files locked, now unlocking lock
> [19725] dbg: locker: safe_unlock: unlocked
> tc/sa/local/.spamassassin/bayes.mutex
Yeah, that's the expiry first pass. As it said, it couldn't find a
good atime delta to use for expiry, so it didn't anything.
"man sa-learn" has a large amount of information about how all of this works.
--
Randomly Selected Tagline:
"Clear with no chance of showers. Temperatures will hold steady at 68�F with
a humidity level of 45%. There is a slight chance of short periods of
darkness followed by frantic scurrying of labstaff personnel. Winds will be
very calm with slight gusts from the SP2 and the Midship Wall."
- http://www.cs.wisc.edu/csl/weather/
Re: (old?) sa hang @ "bayes: can't use estimation method for expiry"
Posted by snowcrash+sa <sc...@gmail.com>.
> > where it just 'sits' for awhile -- no errors in logs -- and,
> > eventually, 'releases' and completes without error.
>
> What are the messages when that happens? It shouldn't just exit.
Here's a recent example,
> sa-learn --force-expire -D
...
[19725] dbg: bayes: can't use estimation method for expiry,
unexpected result, calculating optimal atime delta (first pass)
(waits here ~ 2 minutes ...)
[19725] dbg: bayes: expiry max exponent: 9
[19725] dbg: bayes: atime token reduction
[19725] dbg: bayes: ======== ===============
[19725] dbg: bayes: 43200 221166
[19725] dbg: bayes: 86400 221038
[19725] dbg: bayes: 172800 220701
[19725] dbg: bayes: 345600 220382
[19725] dbg: bayes: 691200 219516
[19725] dbg: bayes: 1382400 217516
[19725] dbg: bayes: 2764800 215351
[19725] dbg: bayes: 5529600 212250
[19725] dbg: bayes: 11059200 210410
[19725] dbg: bayes: 22118400 203758
[19725] dbg: bayes: couldn't find a good delta atime, need more
token difference, skipping expire
[19725] dbg: bayes: expiry completed
[19725] dbg: bayes: untie-ing
[19725] dbg: bayes: files locked, now unlocking lock
[19725] dbg: locker: safe_unlock: unlocked
tc/sa/local/.spamassassin/bayes.mutex
>
Re: (old?) sa hang @ "bayes: can't use estimation method for expiry"
Posted by Theo Van Dinter <fe...@apache.org>.
On Wed, Sep 19, 2007 at 08:34:47AM -0700, snowcrash+sa wrote:
> [19416] dbg: bayes: expiry check keep size, 0.75 * max: 112500
> [19416] dbg: bayes: token count: 218650, final goal reduction size: 106150
> [19416] dbg: bayes: first pass? current: 1190215099, Last:
> 1110223381, atime: 1261902, count: 28215, newdelta: 335417, ratio:
> 3.76218323586745, period: 43200
> [19416] dbg: bayes: can't use estimation method for expiry,
> unexpected result, calculating optimal atime delta (first pass)
> [19416] dbg: bayes: expiry max exponent: 9
>
> where it just 'sits' for awhile -- no errors in logs -- and,
> eventually, 'releases' and completes without error.
What are the messages when that happens? It shouldn't just exit.
> this only seems to happen an unquantified "some of the time" ...
> sometimes (as moments ago) repeating the process results in no such
> hand/delay.
It's running the first pass. See "man sa-learn".
--
Randomly Selected Tagline:
If at first you don't succeed, cry.