Limit of 40 sharepoints in 10.4.x server???

Question

Well, well, well. I'm a bit peaved as I write this. We have been unable to upgrade to Tiger Server, and the big reason is PERFORMANCE. In our environment, Tiger Server simply grinds to a halt when under a normal load (CPU usage at 100% per CPU). This same load on Panther Server consumes merely 25% of our dual G5 server's CPU power (12-13% per CPU).
I opened a case with Apple, and they tell me that they believe the problem is that we are using too many sharepoints (we have 60 defined sharepoints; Apple tells me that they recommend fewer than 40; funny, I don't remember reading that in any of Apple's documentation). Apple's suggestion is to cut down the number of sharepoints. Otherwise, they are unwilling to work with us.
I'm really wondering what might happen if I cut 20 shares from the server over the weekend and see what happens on Monday. I'm thinking my phone might not stop ringing, but that's just a guess.
Apple, we pay good money for your products. We expect them to work. Why in the world would you REDUCE the number of sharpoints available to users? Our Windows server do not have these problems, even when Microsoft releases major updates. Our Windows servers have at least 90 shares, and they don't even break a sweat. We've had a very rocky past with Apple's bugs in their server product, and ExtremezIP on a good ol' Windows Server 2003 system is really starting to look good. Yes, it might require a complete redesign of our server architecture, but, in my opinion (based on past experiences), we would not be throwing our money away (we purchased $1600 in Tiger server licenses that have been sitting on a shelf for 8 months).
Apple (or anybody), any feedback?

Answer 1

hi David...
I did not mention in my previous post that we have TWO Mac OS X servers, one an OD Master and one an OD Replica. The OD master is giving us the problems. The OD replica is fine, and always has been.
However, here are some answers to your questions:
What version of 10.4 have you worked with, there were
known problems under load with AFP in early
iterations of 10.4, all feedback I have seen from
knowledgeable sources is that this issue is greatly
improved as of 10.4.6
We have seen this in every version to 10.4.x, right up to version 10.4.7.
What hardware is in use for 10.3.x server and how
much ram, what hard drive configuration(s), vs. the
same for your 10.4.x server ?
It's the same server.
What do you see in the logs for AFP ?
/Library/Logs/AppleFileService
Nothing significant (trust me...we've scoured them).
Regardless of how the files behave under 10.3.9, how
did you migrate them to your 10.4.x server, and did
you run a Disk check/repair ?
Two ways:
1. I did the upgrade (easy and lazy, I know) to 10.4.
2. I did a COMPLETE reinstall from scratch (format drive, etc) and rebuild of the server. This made no difference.
What kind of differences in network topology (if any)
for the 10.4 server ?
None...it's the same server.
What other services are running on the Server, the
one running 10.4.x ?
Fewer services than are running on the 10.3 configuration (for example, we turned off printing, Macintosh Manager, and SNMP).
Did you clear out .DS_Store files, you might even
want to prevent creation of any new ones (although
the directories will not retain any customization of
window-settings). See
http://docs.info.apple.com/article.html?artnum=301711
We have over 3TB of data, and I don't want to risk messing with the raw data in any user/department folders unless Apple directs me to do so. This is an option, of course, but not one that I'm willing to take at this time. And, remember that our OD Replica works fine.
Have you checked the status of the ethernet port ?
There could be auto-negotiation issues between the
port in the Apple server and your switch. This is not
an "Apple" issue, it happens with Win2k3 servers and
some NICs and switches.
We did all this. We have a 100% new switched architecture resting on Cisco equipment that has been in place for a year. There are no problems. We check this daily. And, I must say, when I reboot the server (*identical computer*) into 10.3.9, the problem go away.
Hard set both ends if need be. Set both to auto-neg
or hard-set both.
Doesn't matter if it's not an issue in 10.3.9, verify
it even if it's matching hardware that's running
10.4.x
See my previous statement. It's the same hardware. There's nothing wrong with the network.
Finally, read this article and perhaps try some of
the tuning(s) suggested:
http://www.afp548.com/article.php?story=20060329213629
494
I already tried this. Nothing here worked.
See my frustration?
I am going to try Apple's suggestion by moving as many of the sharepoints over to the OD Replica as I can. I believe that I can trim the number down to the high 30s on the OD Master.
I will update as I hear more.