Debug on a production server

[Scott Silva]

on 10/12/2007 1:21 AM Julian Field spake the following:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> 
> 
> Scott Silva wrote:
>> on 10/11/2007 1:38 PM Mikael Syska spake the following:
>>> Scott Silva wrote:
>>>> on 10/11/2007 12:02 PM Mikael Syska spake the following:
>>>>> Ugo Bellavance wrote:
>>>>>> Mikael Syska wrote:
>>>>>>> Hi,
>>>>>>>>> There does not seem to be much info on this ... and my scan 
>>>>>>>>> times are also rather high ... not that its a problem atm ... 
>>>>>>>>> but it could be in the future :-(
>>>>>>>> Please provide more information:
>>>>>>>>
>>>>>>>> Hardware
>>>>>>> OS: FreeBSD 7 ( yes its current, but 6.4 did not perform very 
>>>>>>> disk with the SAS 5iR controller
>>>>>>> 2GB ram
>>>>>>> Dual Core Intel Xeon 3060 2.40 Ghz
>>>>>>>> # of child processes
>>>>>>> 8
>>>>>>>> scan times of full batches.
>>>>>>> Oct 11 18:48:58 spam02 MailScanner[72858]: Batch (15 messages) 
>>>>>>> processed in 89.57 seconds
>>>>>>> Oct 11 18:49:08 spam02 MailScanner[72872]: Batch (15 messages) 
>>>>>>> processed in 88.72 seconds
>>>>>>> Oct 11 18:49:10 spam02 MailScanner[72854]: Batch (15 messages) 
>>>>>>> processed in 106.89 seconds
>>>>>>> Oct 11 18:49:19 spam02 MailScanner[72865]: Batch (15 messages) 
>>>>>>> processed in 105.85 seconds
>>>>>> Looks fine.  Is there a reason why you use 15 message batches?
>>>>> you mean instead of 30 ....
>>>>>
>>>>> Some performance turning I read on the wiki ... but t does not seem 
>>>>> to have any effect on my system ... so it will do up to deafult again.
>>>>>>>> Using RBLs at MTA
>>>>>>> nope ... we have had very bad exprerience with that ... both 
>>>>>>> tried spamcop and spamhaus ... both have to many FP here in 
>>>>>>> denmark ....
>>>>>> Spamcop is FP-prone, but I've never heard of a FP in north america 
>>>>>> for spamhaus.
>>>>> Then you are a lucky man ...
>>>>>
>>>>> since the server aint that overloaded I dont see any reason to risk 
>>>>> getting any FP ...
>>>>>>> Its not a problem that I takes so long time .. just saw the 
>>>>>>> message about the patch and wandered if that would make a diff on 
>>>>>>> my scan times ...
>>>>>> Ok, I doubt so. Did you put the MailScanner working dir and /tmp 
>>>>>> in memory (tmpfs on linux)?
>>>>> no ... its on the disk ... and since every mail could be far too 
>>>>> important I dont intend to use it ....
>>>> Tmpfs is absolutely safe on mailscanner if you follow the wiki and 
>>>> only put the mailscanner incoming directory there. And the speed 
>>>> increase is very noticeable, especially in virus and spam scanning.
>>>> Mailscanner does not actually remove any messages. It sees the 
>>>> message in mqueue.in, extracts it to incoming, does its work, and if 
>>>> messages are clean it hard links it to mqueue and then unlinks from 
>>>> mqueue.in. So there is no chance of mailscanner losing a message. If 
>>>> it dies at any point up to the unlink, the original message is in 
>>>> mqueue.in waiting to be processed again.
>>> You mention the wiki ... I can only see 
>>> http://wiki.mailscanner.info/doku.php and a link to: 
>>> http://www.sng.ecs.soton.ac.uk/mailscanner/serve/cache/120.html witch 
>>> does not seem to work.
>>>
>>> and there does not seem to be anything about tmpfs ... if ... then 
>>> I'm not able to find it ...
>>>
>> Julian,
>> Do you have any of this old material ( like 
>> http://www.sng.ecs.soton.ac.uk/mailscanner/serve/cache/120.html) 
>> archived somewhere?
>> I would be willing to spend some time fixing this up if I had the 
>> content to put in.
>> I don't want to go from memory, as I will probably get something 
>> really wonky.
>>
> Thank you very much, it's greatly appreciated!
> The old material is now online again at
> http://www.sng.ecs.soton.ac.uk/mailscanner.archive/serve/cache/
> If you could get it added to the wiki (the useful bits, anyway :-) that 
> would be great.
> 
There is much outdated info in here. I will need to spend more time trying to 
fix it up where I can, and post links to parts that I need experts in that 
area (postfix, exim, sql, etc...) to clean up.
I will also try to leave docs as generic as possibe, and not use full paths, 
since they are different from rpm and tarball installs.

Julian,
Are you going to leave the old cache up for a while, or are you going to 
remove it soon? I thought about fixing links into the old docs at first until 
I can fix the entire section.

-- 
MailScanner is like deodorant...
You hope everybody uses it, and
you notice quickly if they don't!!!!