Re: [AMaViS-user] amavis-milter: (mlfi_eom) smfi_addheader failed

[Mark M. <Mar...@ij...>]

Philippe,

| The problem happens once or twice a day on the previously mentioned hardware 
| platform. We have another mail gateway (single PproIII 1.4GHz + 1Go RAM) with 
| the same software config and no amavis-milter crash during the last week.
| Here is a copy of last night crash report. This one seems to be linked with 
| SA_TIMEOUT instead of smfi_addheader.

| avr 14 03:29:56 atlas amavis[7134]: (i3E1TOf3019897) SA TIMED OUT, backtrace: 
|   at /usr/share/perl5/Mail/SpamAssassin/BayesStore.pm line 573\n\teval {...} 
| 
| avr 14 03:35:24 atlas amavis[16911]: (i3E1ZHFs009868) Sophie av-scanner 
|   FAILED: Too many retries to talk to /var/run/amavis/sophie.sock (Can't 
|   connect to UNIX socket /var/run/amavis/sophie.sock: Aucun fichier ou 
|   r\351pertoire de ce type) at (eval 37) line 180.
| 
| avr 14 03:37:40 atlas amavis[16911]: (i3E1bc33000204) Sophie: Can't send to 
|   socket /var/run/amavis/sophie.sock: Noeud final de transport n'est pas 
|   connect\351, retrying (1)

It looks like the system was very busy during that period,
both the SA and Sophie were failing.

Since the BayesStore.pm is mentioned in the traceback, it may be
worthwhile to stop amavisd and rebuild the SA Bayes database
(su vscan; sa-learn --rebuild --force-expire)


| Apr 14 03:39:26 atlas sm-mta[19897]: i3E1TOf3019897: Milter: data, reject=451 
|   4.7.1 Please try again later
| Apr 14 03:39:26 atlas sm-mta[19897]: i3E1TOf3019897: Milter (milter-amavis): 
|   timeout before data read
| Apr 14 03:39:26 atlas sm-mta[19897]: i3E1TOf3019897: Milter (milter-amavis): 
|   to error state
...
| Apr 14 03:40:36 atlas amavis-milter[899]: (mlfi_cleanup) unlink failed: Bad 
|   address
| Apr 14 03:56:39 atlas sm-mta[9037]: i3E1kcCs009037: Milter (milter-amavis): 
|   init failed to open

Was the process amavis-milter still alive but stuck after this event,
or did the process die?


In case you have already fetched the amavis-milter.c from our web,
would you please so it again:

  http://www.ijs.si/software/amavisd/amavis-milter.c

I added reporting of queue-id to log messages, so that it will be
more obvious which log entry pertains to which message
in case that several messages are being processed in parallel.

  Mark