FreeBSD amd64 and compatibility

S

soylentgreen

Our mail server has been chewing up the swap file. We used to run FreeBSD 7.0 i386 with a PAE kernel with 8gb RAM, and it worked like a charm. When I upgraded to 7.1, the PAE kernel locked up several times per day and destroyed the user folders on the RAID array; at that time it was recommended to me that the PAE kernel was not a good idea.

I had heard about some potential incompatibility with hardware / applications with the amd64 bit version, so the company decided that with the amount of time we had to work with it would be easier to just do a clean install of i386 and only use 4gb ram. Now here we are with a system that eats about 1% of the swap file every day, so after about 50 days, its half full and I get paranoid and reboot the server.

A RAID controller recently lost a Seagate drive (big surprise) so I put the backup server into place, and now I have the main mail server offline and was given the go-ahead to try the amd64 version so we could again use all 8GB of the RAM. I have questions regarding potential incompatibilities on the hardware and software if someone would be so kind as to enlighten me.

(hardware)
Motherboard = Supermicro X7DBN
CPU(s) = 2x Xeon 5410 Quad Core
RAID controller = 3ware 9650SE-8LP
with BBU-Module-03

(software)
Sendmail
(ports)
Procmail
Dovecot
Apache
The Horde (Groupware)
Imp, Dimp, Kronolith, Turba, etc.
Spamassassin
ClamAV
LDAP
and a multitude of dependency ports.

There is no Xorg or Java or anything like that running on this server. Is anyone running anything similar to this with the 7.1 or 7.2 series of amd64, or does anyone have any ideas as to any problems I might have with the above hardware/software? (would I have to do NOAPIC or disable APCI or anything?)

Thanks for your attention.
 
soylentgreen said:
Our mail server has been chewing up the swap file. We used to run FreeBSD 7.0 i386 with a PAE kernel with 8gb RAM, and it worked like a charm. When I upgraded to 7.1, the PAE kernel locked up several times per day and destroyed the user folders on the RAID array; at that time it was recommended to me that the PAE kernel was not a good idea.
Click to expand...

PAE kernels aren't ideal, but they should work fine. It would be nice to track down what went wrong with the 7.1 PAE kernel. Perhaps try 7.2 to see if it's still there?

soylentgreen said:
I had heard about some potential incompatibility with hardware / applications with the amd64 bit version, so the company decided that with the amount of time we had to work with it would be easier to just do a clean install of i386 and only use 4gb ram.
Click to expand...

Generally there shouldn't be any issues with the 64 bit version.

soylentgreen said:
Now here we are with a system that eats about 1% of the swap file every day, so after about 50 days, its half full and I get paranoid and reboot the server.
Click to expand...

This would trouble me. It sounds like the software you are using is leaking memory. Your decision to reboot is probably unnecessary, but finding the leaking process and restarting it should be just as good. It's course wise to prevent the system from running out of memory. It would be good to find the source of the memory leak and fix it.

soylentgreen said:
A RAID controller recently lost a Seagate drive (big surprise) so I put the backup server into place, and now I have the main mail server offline
Click to expand...

Didn't replace the drive and have the RAID rebuild? Why not? Just wondering...

soylentgreen said:
and was given the go-ahead to try the amd64 version so we could again use all 8GB of the RAM. I have questions regarding potential incompatibilities on the hardware and software if someone would be so kind as to enlighten me.

(hardware)
Motherboard = Supermicro X7DBN
CPU(s) = 2x Xeon 5410 Quad Core
RAID controller = 3ware 9650SE-8LP
with BBU-Module-03
Click to expand...

I don't have any experience with this specific hardware, but do have a couple Supermicro systems (PDSM4+), and they work fine. I doubt you'll have problems, but of course can make no guarantees.

soylentgreen said:
(software)
Sendmail
(ports)
Procmail
Dovecot
Apache
The Horde (Groupware)
Imp, Dimp, Kronolith, Turba, etc.
Spamassassin
ClamAV
LDAP
and a multitude of dependency ports.


There is no Xorg or Java or anything like that running on this server. Is anyone running anything similar to this with the 7.1 or 7.2 series of amd64, or does anyone have any ideas as to any problems I might have with the above hardware/software? (would I have to do NOAPIC or disable APCI or anything?)

Thanks for your attention.
Click to expand...
 
[cmd=top]-o res[/cmd] might help you find processes that have a memory leak, because that will sooner or later be the process that reserved most memory. And end up at the top.
 
swills said:
P
Generally there shouldn't be any issues with the 64 bit version.

This would trouble me. It sounds like the software you are using is leaking memory. Your decision to reboot is probably unnecessary, but finding the leaking process and restarting it should be just as good. It's course wise to prevent the system from running out of memory. It would be good to find the source of the memory leak and fix it.
Click to expand...

Just to follow up:

1, I installed the 7.1 amd64 bit version, and I don't seem to have any problems, hardware or software; I chose the 7.1 version as I had read the life-cycle on this version was longer than the 7.2 version. I've put this thing through the wringer as far as stress testing and sending / scanning messages. I had ended up upgrading to all the latest ports so hopefully the 'bad' program that was eating the page file is no longer an issue. I changed the clamav-milter to run from the clamd daemon with 0.95.1 (which is the way you have to do it now anyway), and it seems to be using about 70% of the memory resources it was before.

swills said:
Didn't replace the drive and have the RAID rebuild? Why not? Just wondering...
Click to expand...

I wanted to get to the bottom of the potential memory leak issue. At the time the server went down, I wasn't sure if the RAID drive going bad was the only hardware issue, and I wanted to add the extra ram.

Things appear to look good. My neurotic testing and backups are ready, and I'll be putting the 'big' one back into place maybe tomorrow morning or Friday.

Thanks for your opinions.
 
You must log in or register to reply here.
Back
Top