Yes - but how many users (on a typical ISP) will this support?  My
experience with NFS suggests that the number is less than 10000 accounts.
In an early trial, reguardless of load, checking a POP account required
10-15seconds each.  Normally, the same check is instantaneous on local
drives.  On the pop server, we normally have 5-10 simultaneous sessions on
POP with about 200 per minute.  On NFS, we could only process 100
connections per minute and the number of simultaneous connections jumps to
40+.  After several minutes, the simultaneous conns started exscalating and
the server crashes at 500.  The same machine has no problem on local hard
drive even during heavy spammings.

Duane.

-----Original Message-----
From: Gene Parks [mailto:[EMAIL PROTECTED]]
Sent: Monday, March 11, 2002 3:01 PM
To: [EMAIL PROTECTED]
Subject: RE: high availability solution?


If you had a common storage area i.e. a NAS then you could just bring
the second server online mount the NFS and add the mx record.  This
would accomplish what you are looking for since when you take one out of
service the other mail server trying to send mail will use the second
record. And since the mail is local to all boxes then you do not have to
worry about the mailhost attribute or clustering.

-----Original Message-----
From: manfred [mailto:[EMAIL PROTECTED]]
Sent: Monday, March 11, 2002 6:52 AM
To: [EMAIL PROTECTED]
Subject: high availability solution?


Hi everybody,

I have a little understanding question to qmail-ldap. I haven't found an

answer to this in the documentation yet, so I'll give it a try here.
I'm running qmail-ldap without clustering enabled and it works really
nice.
Now I want to have a second server installed for the case, that the main

server has to be maintained or is going down. As I understand (please
help me
if I'm wrong here which I certanly might), qmail-ldap has clustering
support
which is designed for load-balancing in the first place. So, when
cluster-members receive pop3 sessions, they try to hand it over to the
mailHost cluster-Server, right? But what, if this one is down? Does the
session end with an error for the pop user?
I assume, that smtp sessions (local delivery) will work and the
cluster-member tries to connect to the mailHost cluster when it's up
again to
sync the maildirs, or am I wrong?
So another way would be, to have a second MX entry, which would be fine
for
'hot-standby' of the second server, but let's assume the case, that a
lot of
mails get handled locally in the time, the first server is down. Then,
how is
it possible to have the mails delivered from the second server to the
first,
so the maildirs of the first server have all the mails (because ongoing
pop
sessions will be handled by the first MX again)?
I would really appreciate some hints on this topic, maybe a good
starting
point in the web for further research, befor I try to configure
'something'.
Thanks a lot.

Manfred

Reply via email to