Performance question

[Mariano Absatz]

Do you know what kind of filesystem is /var/mqueue.in on? (e.g. ext2).

The standard unix filesystems (including ext2 and, I think ext3 also) are 
_reaaaaallyyyyyyy_ slow on very large directory searches.

Add to that the fact that MailScanner tries to order queue files by date 
(stat-ing the files in the queue for every batch).

Long before investing in new hardware I would:

1) install the new MS 4.24 which has an emergency mode triggered when the 
input queue exceeds a (configurable) treshold:
Max Normal Queue Size = 5000
(you can modify this value, but don't set it too low).
This will turn "emergency mode" on when it finds the input queue has more 
than 5,000 messages.
During "emergency mode", MS won't try to order queue files and will 
process as it finds them.
This should:
a) increase (significantly) the overall MailScanner throughput
b) eliminate (during that period) MailScanner "fairness", since old 
messages may be processed well after newer ones.
MailScanner will stay in "emergency mode" for 40 more rounds _after_ the 
queue dropped below "Max Normal Queue Size" before switching back to 
"normal mode" in order to try to completely drain the queue and not have 
to be switching modes constantly.

2) the other thing you can do is to switch the filesystem type (using 
ReiserFS, or other of the new FSs... however doing that is well beyond 
the scope of this list).

Regards.


El 8 Oct 2003 a las 11:00, Alan Fiebig escribió:

> Hello,
> 
> I'm currently running MS 4.23-11 with SA 2.55, Bayes disabled, RBL
> disabled, no offsite checks (pyzor etc.) in use. 
> Sendmail is my MTA.
> 
> I am using tmpfs for my MailScanner/incoming directory.
> 
> On occasion I get BIG backups in my mqueue.in
> 
> When this happens, MS is indeed processing messages, but exceedingly
> slowly. But yet my server seems to have tons of resources available, the
> server is not running out of steam. Rather, most of my MS children are
> sleeping. 
> 
> here is an example:
> TOP shows 17 MailScanners sleeping, 1 running
> uptime shows CPU usage of 1.86%, 3.38%, 5.78%
> Zero swap space is in use
> Approx 1GB of RAM is inuse, 200KB RAM free
> 80% CPU idle time
> 
> 9,000 messages are in mqueue.in
> 902 in MailScanner/incoming
> 
> Watching the log shows no timeouts, and over an 18 minute period the
> server processed 2,200 messages, or 7,300/hour, 176k/day 
> As we receive closer to 250,000/day, this just doesn't cut it. Yet I see
> no indication that faster or more hardware would have any impact
> considering the server is basically resting. 
> 
> Any ideas? Thanks!
> 
> -Alan


--
Mariano Absatz
El Baby
----------------------------------------------------------
I've never met a human being who would want to read 17,000 pages
of documentation, and if there was, I'd kill him to get him out
of the gene pool.
                   -- Joseph Costello, President of Cadence