Hash or btree for access.db

mail

    Next

  • 1. Configure sendmail to add BCC to all sent mail..
    Hi to all, we're using sendmail as smtp daemon on our company's mail server. Some employees use a special account configured on their mail client software (e.g XXXX@XXXXX.COM ) to answer costumers' requests. We'd like to feed a database with all sent mail (the sender is obviousely keept anonymous, since the account used to send mail is the same for all employees): in order to do this, we asked the employees to add a bcc to a special mail address. A daemon, running on another server, periodically reads the mailbox asssigned to the special mail address and elaborates messages. The problem is, in the most of answers the employee simply forgets to add a BCC, so we're searching for a method, or plugin, to make sendmail add automatically the special BCC address to all mail sent from XXXX@XXXXX.COM . There's a way to do this ? I understand that a such method may be used to violate employees' privacy, but this is not the case, 'cause it would be applied only to a non-personal mail account and employees will always be aware all sent mail 'll be recorded. Thanks in advance, and sorry if this's a FAQ. Scorpio.
  • 2. delivering time
    I use Mutt with Sendmail as MTA. If I send an e-mail when I'm not connected to Internet, it occours more or less a minute for the e-mail arrives to its spool directory, i.e. /var/spool/clientmqueue. Is there a method, a string to put in submit.cf, to shorten this time? Is all regular, that an e-mail needs a so long time to be put in a dir in the same pc? Thanx M
  • 3. Chokes
    REcently, some of the Big Mail Servers like yahoo ang mail are choking sending e-mail to our servers. Recently the mail server was subject to s DDoS SYN Floof Attack. What can I do to unchoke this. -- Member - Liberal International This is XXXX@XXXXX.COM Ici XXXX@XXXXX.COM God Queen and country! Beware Anti-Christ rising! Beware Linux the Microsoft of Unixes

Hash or btree for access.db

Postby Ray Abbitt » Thu, 16 Jun 2005 14:40:36 GMT

Ok, I've done some searching and can't find any definitive answer as
to which is best. The only differences I can see on my (old slow)
system is that makemap hash takes over 5 minutes to run and makemap
btree takes just a little over 1 minute. (My access text file has 
about 100000 lines). Am I taking a performance hit with sendmail if
I use the btree rather than hash? I've tried it both ways and can't 
really see any difference other than the time it takes to rebuild the
database.

-ray

Re: Hash or btree for access.db

Postby Andrzej Adam Filip » Thu, 16 Jun 2005 15:14:00 GMT



Btree is better for building BIG dtabases from "key sorted" input files.

Using big access db is not recomended. Sendmail can do 20-30 access 
lookups per one incoming message. When big part of access.db file is not 
cached by library itself or file system cache it may create big 
penalties approaching one disk I/O operation per one lookup.

-- 
Andrzej [en:Andrew] Adam Filip  XXXX@XXXXX.COM   XXXX@XXXXX.COM 
"All that is necessary for the triumph of evil is that good men do
  nothing"  -- Edmund Burke (18th century)


Re: Hash or btree for access.db

Postby Ray Abbitt » Fri, 17 Jun 2005 02:40:47 GMT

In article <d8oguj$541$ XXXX@XXXXX.COM >,




But how big is big? Is 100000 or so entries considered big? And for
what it's worth, my input files are (mostly) sorted.


Once again though, how big is big?  My access.db is around 16 Mb (it is
about 30% larger using hash instead of btree). I haven't seen any
performance differences, but the system isn't particularly busy either.
There are only a few real users and message traffic is in the thousands
per day. It's a PII 266 with 128 Mb of ram and load average tends to run 
around .05 or less most of the time. 

-ray


Re: Hash or btree for access.db

Postby Andrzej Adam Filip » Fri, 17 Jun 2005 03:22:00 GMT






It depends on memory size (size of in memory file system cache).
IMHO 1E5 is not a big db for relatively modern PC server.

Building btree from sorted (or partially sorted) input file reduces 
number of required disk I/O operations when created db is not fully 
cached by system cache of file system.


Can you report the following:
1) (peak) physical ram usage [size of free/available physical memory]
2) number of messages received in the busiest hour of week
3) use of AS/AV software (scanning message bodies "in smtp session")

Use some software to track memory usage, system (CPU) load and disk I/O 
operations (It is OS/distribution dependent) - it should provide you 
more reliable indicators.

-- 
Andrzej [en:Andrew] Adam Filip  XXXX@XXXXX.COM   XXXX@XXXXX.COM 
"All that is necessary for the triumph of evil is that good men do
  nothing"  -- Edmund Burke (18th century)



Return to mail

 

Who is online

Users browsing this forum: No registered users and 71 guest