Re: [EXIM] MFS mount spool

Kezdőlap
Üzenet törlése
Válasz az üzenetre
Szerző: Richard G. Duvall
Dátum:  
Címzett: Tabor J. Wells
CC: exim-users
Tárgy: Re: [EXIM] MFS mount spool
We are using a narrow SCSI controller, but a Wide drive (barracuda). That
drive is where we store the mail (/var/mail), and that is the drive that
has the root filesystem.

However, we have a seperate drive that is an older drive that is a narrow.
I put the spool on this. It significantly improoved the speed.

I would like to see the root filesystem on a seprate drive so that when
mail is being delivered into their maildrops, it will be happening on a
seperate drive as the password file, because it takes forever to
authenticate sometimes. Would be nice to mirror the /var/mail drive.
don't know what to do about the spool, though, wether to mount in memory,
or to get a faster drive.

so, in our situation, what do you suggest?

Get a wide controller, swap our our 9g barracuda with a 9g cheetah, put
root filesystem on different drive?

This is getting interesting. At what point do we need to have 2 mail
servers and do load balancing?

Sincerely,

Richard G. Duvall

On Thu, 17 Jun 1999, Tabor J. Wells wrote:

> On Thu, Jun 17, 1999 at 03:50:05AM -0700,
> Richard G. Duvall <rgduvall@???> is thought to have said:
>
> > Yes, but the drive is cranking like mad! Should I install a different
> > drive on a different SCSI bus just for spooling?
> >
> > It's even making authentication slow, and it is running on a Pentium II,
> > 400Mhz with 128MB Ram.
> >
> > It's just unbearable the amount of time it takes to authenticate. Alot of
> > time when I am running PC-Pine over the same ethernet, it times out. this
> > is really annoying...
> >
> > I tried installing a different drive just for spooling on the same scsi
> > bus, and this helped tremendously. But, I still get timeouts with
> > PC-Pine, and iostat is reporting TONS of activity on both drives (spool,
> > and the other drive which contains root filesystem, and where mail files
> > are being stored).
> >
> > Our weakest link seems to be the speed of our hard drives, and it is
> > slowing authentication down. What can we do to remedy this problem? We
> > are running that 50 pin scsi card from BusLogic, and one of those 9 gig
> > Segate drives, in which we have to use an adaptor thingy to make the cable
> > work on it (counted 68 pins on the drive). Then, the other drive where I
> > am spooling the stuff is 50 pin, and the cable plugs right into it.
> >
> > Do I need to get one of those faster SCSI cards from Adaptec that plug
> > directly into that kind of drive without an adaptor thingy. Would that
> > make it faster?
> >
> > Sincerely,
> >
> > Richard G. Duvall
>
> Ok. It sounds like you've taken a fast/wide SCSI drive and plugged it into
> a narrow SCSI adaptor. That's going to significantly affect the speed of
> your disks. Also you say it's a 9g Seagate, which kind? Is it a Barracuda
> or Cheetah? Barracudas spin at 4000 RPM, Cheetahs at 10,000 RPM. I
> wouldn't use a 9g Barracuda in an I/O intensive environment. It just takes
> too long to seek across the drive.
>
> Also you can improve your performance quite a bit but locating your spool
> on a separate filesystem. It doesn't necessarily have to be a separate
> controller, but that wouldn't be a bad idea if you're doing a lot of mail.
>
> At Shore what we've got is two 4g drives stripped together and then
> mirrored (RAID 0+1) using RAID software (Sun's Solaris Disk Suite). I
> don't know what options like that exist for you in your OS of choice, but
> you might look into doing something like that as well, if necessary.
>
> Tabor
>
> -- 
> ___________________________________________________________________________
> Tabor J. Wells                                             twells@???
> Systems Administration Manager  Just another victim of the ambient morality
> Shore.Net  --  High quality Internet access and hosting services since 1993

>
> --
> *** Exim information can be found at http://www.exim.org/ ***
>
>



--
*** Exim information can be found at http://www.exim.org/ ***