First I have to apollogize that this posting is not 100% XMail 
related.

Having said that, my problem is as follows:

  We run XMail v1.9 on an Intel PIII 900MHz with 256MB RAM, using 
Slackware Linux 8.0 Kernel 2.4.9 for about half a year now with one 
domain and about 70 users. Traffic is about 30 MB of mail per day 
(sorry, since I do no logging I have no Idea of messages per day) 
ranging from 15 to 70 MB/day.

  So what's the problem ? Simply performance. But not allways. For 
months the system worked _very_ fine, responses where fast and 
transfer rate only limited by the connection's bandwith. And then, 
slowly the throughput eroded, sometimes coming to a grinding halt.

  Well what happens ? If users try to download mails, their client 
starts loading messages at a _realy_ low speed (a minute or two for 
some plain text message containing 20 lines pf text or so) and 
eventually times out. At the same time the load at the XMail box is 
about 0.01 and there is no network congestion as far as I can see. If 
they try half an hour later, everything is fine, or not, you never 
can tell.

  I tried to push data across the router in front of the mailserver 
by copying a large file via scp from my destop PC to the mailserver: 
transfer rate was close to the theoretical maximum of the network 
segments used (10 MBit Ethernet hub between the router and the 
mailserver) and system load at the router (also Linux but a Pentium 
MMX 233) was about 0.25.

  So, the machines and the network are capable of far more than I 
need, where could the bottleneck hide ?

  Last Friday I found estematedly 30 frozen messages, which I deleted 
after a very brief review (they all suffered misstyped domain or 
usernames) and users reported an improved situation on monday. But 
today there are just 3 frozen messages and we are back at snail mail 
again ;-)

  Does anybody of you have any bright idea what to do about this 
situation ?

  I can supply some further observations: 

* System uptime is 120 days now (but that usually is no problem in 
the UN*X world and I never saw Linux boxes suffering from long 
uptimes, unlike ... no, I will not start a new flame on that ;-) 
....). 

* Sometimes, when I dig around in the system via ssh from the company 
lan (so only the router between me and the mailserver) I can watch 
the characters I just typed in appear on the screen one after the 
other, quite slowly (again something I am NOT used to)

* I use my own perl virus filter script together with ClamAV - so 
sometimes, when some stupid sends a 5MB attachment to all users 
(about 70) the system gets quite busy. System load 16 (!) for an 
hour, scanning the same file 70 times, simultaneously. But it keeps 
responsive all the time ... slowly, but you still can mail and work 
on the box (great, those UN*X systems, aren't they ?) ... by the way, 
this is definitely NOT the problem I described above, because it only 
happens every now and then (users are not that silly ;-) )

* I watched with 'netstat -cteo' ... I see two to six simultaneous 
connections at best ... when someone has the 'timeout problem' the 
'Send-Q' column in the output of netstat goes up to about 15000 which 
looks quite normal, compared to my scp-performance tests.

* DNS works great ! All client connections resolve within a moment's 
notice, local as well as remote.

  I am quite puzzled. Any help _very_ welcome !

  Goesta


-- 
Let's stand together against 'Palladium' ... 

Link arms, don't make them
http://online.securityfocus.com/columnists/96

-
To unsubscribe from this list: send the line "unsubscribe xmail" in
the body of a message to [EMAIL PROTECTED]
For general help: send the line "help" in the body of a message to
[EMAIL PROTECTED]

Reply via email to