Hello,

I use Logwatch and it daily mails me about what happened on the server a day ago.
I have SpamAssassin with Amavis on Ubuntu Server with Dovecot and Postfix email software.

The first error that shows me is this: "spamd internal error, python traceback seen in response" - what is this about ?
The second error that shows me is this: "idle children more than 2 maximum idle children. decreasing spamd children:"

What can I do to resolve these ?

And most importantly: we receive and send a lot of important mails !
Can any of the above errors lead to the emails not being delivered ?

Thank you !

Has Spam Assassin migrated to python? Thank goodness.

spamd/spamc is a server/client pair. One (Probably spamd) is exiting on the python tailspin by the sound of it, and leaving orphaned processes. You have a mess on your hands.

Yes, important emails could go awol, and you need to spend time there. The mail system pipes mail along the chain, and if one of the pipes is broken …

I found it useful to have a test mailbox of spam and cat it at 127.0.0.1:25 or somesuch (It was a long time ago). You should get all your test mail in the spam, all things being equal.

Get digging, and find out what's going on.

Cross-posted @Ubuntu Forums

I already tested it by sending 32 emails in a matter of 1-2 minutes, enough to trigger that child kill message again.

However, all 32 emails were delivered with no problem.

Are you absolutely sure that emails are dropped?

Would it help if I posted the email server configurations or a basic schema of my layout?

No, don't post configs for me anyhow. My experience was 20 years back, SpamAssassin was in Perl, and I don't need reading material.

What's "enough to trigger that child kill message again?" Is there a resource or loading issue here? I ran email on a puny box (AMD i586 = a '486 with Pentium instructions added, iirc) and laughable memory by today's standards, I could fire ≅100 emails at it together and never had resource issues. My chain was
Incoming -->Vipul's Razor --> DCC --> SpamAssassin with procmail passing the parcels. Postfix was doing something too …

Find a way of trapping that spamd error. Your should not lose child processes and you've no way of telling where the data it's handling goes. The fact that you got 32/32 doesn't mean you'll get 100/100. It's your job to do this, by the sound of it.

I also have this problem: " spamd: child [___] killed successfully: interrupted, signal ___ (___): 2 Time(s)" ... what could be the problem ?

Am I the only one here? spamd killing a spamc child isn't necessarily a problem - every message gets a spamc child.