Internal quota calculation error (redis), double mail delivery

[Tom Sommer]

On 2016-04-06 21:22, Timo Sirainen wrote:

>> Yea, but why is it taking 30 seconds anyway? I'll try and lower it, 
>> but I suspect I'll just get 10.000 secs errors instead?
> 
> I don't know. Possibly a hanging TCP connection. That's why I was
> asking about the firewall. If it's not that, I don't really know. And
> yes, just shrinking this without the below fix won't help anything.

Indeed. I have antivirus on the same server doing redis to the same 
redis-server, with no errors.

>> What about the issue of lmtp saving the mail, but sending an error 
>> back to Director?
> 
> I guess the LMTP proxy_timeout in director is 30 seconds or lower. If
> backend's timeouts are smaller than that, director won't have to
> disconnect LMTP due to timeout and there won't be duplicates. Or just
> increase the director's proxy_timeout to several minutes.

proxy_timeout is not set, I'm unsure of the default. I've raised it to 
120 now.

This is the log from director:

Apr 06 15:42:00 lmtp(46187): Error: xxxx: Failed to send message to 
<xxxx at xxx.xxx> at xxx.xxx.com:24: 451 4.4.0 Remote server not answering 
(timeout while waiting for reply to DATA reply) <5vWKIVUIBVdrtAAA1VGibg> 
(1/1 at 30067 ms)

So yea, guess the timeout was 30 there as well.