no, you may be experiencing hardware issues.. will your upgrade include new hardware?

Yes, we have actually ordered new hardware since it seemed the hardware could also be part, if not all of the problem.

I think our engineer was looking for possible hardward loads that seem that people are using that seem to be stable. Like using a 64 bit Centos load versus 32 bit. If either one of those have had issues on specific hardware, but stable on other hardware...etc.

i always use 32bit. i got burned pretty bad with crashes on a 64bit system that the community insisted didnt exist... soon as I reinstalled on the same hardware with 32bit my crashes went away. Plus a lot of licensed modules are 32bit only.

Thanks Erik, that info does help. Can I ask what hardware you are using?

it depends on the application.. for STE we have some customized supermicro hardware we did load testing on with calls in order to certify they work up to a set volume.

for MTE you might want more ass than that. The top of the line option is those HP DL-380 generation 6 servers. They use xeon 5530 or faster chips (quad core + HT), they have that quickPath circuit so they stay off the FSB when talking to each core, and the network controllers have a TCP/IP offload engine. They arent cheap though. Expect to pay north of $6k for them.

however, they have been the only hardware thus far to achieve 10000 call legs (aka 5000 calls).

Do have any suggestions on how much RAM we should have since we can only do 4G on a 32 bit platform?

Eric

youve been misinformed, the PAE kernel allows for more than 4gig

as an example

[root@lou03 modules.d]# uname -a
Linux lou03.xxx.xxxxxxx.xxx 2.6.18-164.15.1.el5PAE #1 SMP Wed Mar 17 12:14:29 EDT 2010 i686 i686 i386 GNU/Linux

[root@lou03 modules.d]# free -m
total used free shared buffers cached
Mem: 8118 1221 6897 0 184 868
-/+ buffers/cache: 168 7950
Swap: 1983 0 1983
[root@lou03 modules.d]#

I guess I was misinformed indeed, thank you.

In terms of doing our upgrade, we are currently running on Thirdlane 6.0.1.78 which is running on Centos 4.7 OS. The new hardware is going to be running on Centos 5.3 OS
My concern in doing this upgrade is that I was planning on just doing a restore from backup to put the exact same config onto the new server. But I'm concerned about the possibilities of not having everything work in the same manner as it was on 4.7.
Should that be a big concern?

Erik:

I used the TL ISO to do an install on one of my servers... then did a yum install Kernel-PAE... I had to edit the grub.conf file to get it to select the new kernel and I had to update/compile the DAHDI drivers. It seems to be working and running without errors. Is there anything else that needs to be recompiled for the new kernel?

It is very nice that it sees all the memory in the server now. Big Question: How much memory should I install? PAE goes to 16Gig... I have enough on hand to go to 8Gig or 12Gig... This is a dual Opteron machine that we would like to handle 100 simultaneous calls or so.

Thanks

if you update dahdi you have to rebuild asterisk from 'make distclean' all the way to make install for it to find the new dahdi header files for chan_dahdi.so

8gig should be plenty

It almost sounds like I shouldn't be able to use the PRI. But I know I didn't go that far and I was still able to make calls... odd.

I thought 8gig sounded good too.

Thanks.

download and compile libpri before re-compiling asterisk