[jpbuse]

Have you had any luck getting this to work? I’ve got a Cisco PIX 515 firewall in place and my remote VPN users are able to do Time Machine backups over the VPN back to the xServe in the main office. I *just* upgraded to 10.5 this weekend and saw this functionality working. I’ll be changing the firewall to NOT allow this as backups from remote users back to the corporate network will be brutally slow.

My VPN users are in a different subnet than my corporate network which makes it easy to create custom firewall rules.

[jpbuse]

[QUOTE][u]Quote by: MacTroll[/u][p]

Sounds like you’re running into drive formatting issues. You can’t do the FWTDM method as it would be APM formatted and the installer, when booted from an Intel box, won’t go onto that.

The current boxed copy of OS X Server is 10.4.6 Universal, so you’d only have to buy the box and not the Xserve… although that won’t be any different from what you’re getting from the ADC.

I’m guessing you either have a G5 Cluster Node or you only have the CD-ROM drive in it, hence the need for the target disk mode? Put one of your laptops into target disk mode and then connect it to the Xserve. Insert the DVD that you burned from the ADC image into that and it will show up on the Xserve.

FWIW, I have 3 G5 cluster nodes in my lab that have all been installed from the universal image. They were all done from FW drives that had the ADC image installed onto them.[/p][/QUOTE]

Joel,

The xServe G5 is not a cluster node and it does have DVD-ROM drive. I used Disk Utility to burn the .dmg I downloaded from ADC if that makes a difference. I didn’t purchase the 10.4.6 univ server version but am using the 10.4.7 version from ADC. The only reason (as of now) that I am trying to get Universal installed on the G5s is due to the issue I’ve outlined in this thread regarding the kernel memory issue(s).

Jason

[jpbuse]

I’m glad my posted fix has helped you guys… unfortunately I am still having issues 🙂 The Universal version does fix this problem, however there is no Univ OSXserver available other than what ships w/ the Intel xServes. I made a bootable copy of one from ADC and also used one from an Intel xServe. While I was able to get the Univ server installed on a xServe G4, I can’t get it to install on my production G5 xServes. The installer will kernel panic during boot.

Besides breaking down and purchasing an Intel xServe does anyone have any suggestions? I tried running the installer from an Intel machine with the xServe G5 in target disc mode but the installer won’t allow me to select that drive. I am going to try using an Intel machine and install to a spare HD, then close the HD for xServe G5.

Now, all of this wouldn’t be necessary if Apple fixed the PPC version. I am running 10.4.10 and so far this issue still persists.

[jpbuse]

[QUOTE][u]Quote by: nightwol[/u][p]Does anyone know if it’s possible on PPC systems (where there doesn’t seem to be an rc.server) to put the kern.maxnbuf=20000 in /etc/sysctl.conf?[/p][/QUOTE]

No, its not. Apple is still working on this issue with me for my PPC machine.

[jpbuse]

Awesome! It doesn’t work on PPC installs but they’re working on a fix for me.

[jpbuse]

Called Apple Ent. They said they have a fix for this issue. The guy I spoke with today was very nice and knew about the issue. You need to edit the rc.server file to release the ram after the find is performed. This file only exists on Universal installs but the tech told me to copy the file from one of my Univ servers to my non ones and it should work just fine.

1. Make a copy of /etc/rc.server.

$ sudo cp /etc/rc.server /etc/rc.server.bak

2. In the text editor of your choice, make the following change to /etc/rc.server. You will need to edit this file as the root user.

Locate this line in /etc/rc.server:

sysctl -w kern.maxnbuf=90000

Change the line to read:

sysctl -w kern.maxnbuf=20000

Save the changes you have made to /etc/rc.server.

3. Verify the changes were made to /etc/rc.server

4. Reboot the server.

[jpbuse]

Is your AFP server tweaked at all? I’ve tweaked my settings according to Nigel Kirsten’s posting at [url]https://www.afp548.com/article.php?story=20060329213629494[/url]. Mainly, my max thread count is 400 and looking at the AFP server right now, it shows 402 threads. I am seeing the CatSearch in the logs again, and AFP is dog slow, however ‘kernel_task’ appears to be consistent… right around 335M for RSIZE and 1.7GB for VSIZE. The machine as 8GB ram total. I checked this against another server also running 10.4.8 but with way less users and so far, no problems and its about the same.

Kernel_task runs around 5-10% cpu as well.

Another issue is after I perform a find on the mounted volume, even if I cancel the fint, I can’t unmount the volume (says its in use) and the CatSearch crap continues to show in the logs until an AFP server reboot.

And yes… to me this is a major major issue. The reason I have the xServe in this case is for AFP and AFP alone. If it can’t do that really well that is a bad bad thing.

[jpbuse]

Update: This morning I checked the AFP logs on my server and noticed the CatSearch stuff again. As always it is coming from one of my Studio users logged into the Studio afp share. Only one machine was creating the CatSearch in the logs and when I checked the machine I noticed it was running InDesign 4.0.0. In the process of updating that to InDesign 4.0.4. So far, none of my machines running ID 4.0.4 have been registering the CatSearch in the afp logs. My guess is now that the offending machine has been updated, the CatSearch in the logs will be gone.

With that said, I am not 100% certain the slowdown of the AFP server is directly related to the CatSearch stuff although it would seem that way. Processes CPU % and memory usage on the server are normal as of now.

[jpbuse]

I am running Retrospect client on the server, not the full app.

Looking at the logfiles I noticed “CatSearch starting [share point]” only seems to show up for certain users of thet server. Actually, only those people in my Studio department. I have other departments using this server.

As of right now, the log shows two machines registering the “CatSearch” stuff. They are both G5 stations, one is running 10.4.7 and the other is running 10.4.8. The users have been using InDesign v4.0.2 all day. Both machines are bound via LDAP. Both accounts are local user accounts.

Update: These are the only machines that had outdated versions of InDesign running on them. My Studio users use ID all day long. All my users were at ID 4.0.4 w/ the exception of the two machines that were showing up w/ the CatSearch in the logs.

I just restarted AFP on the server so nothing to report as of yet. I’ll monitor memory usage of Kernel Task as well.

[jpbuse]

[QUOTE]I first thought it was retrospects fault but…[/QUOTE]

I’m noticing what seems to be a correlation between Retrospect and my server slowdown. I can’t yet confirm anything but looking into it. Do you have Retrospect client installed on your server or are you running Retrospect from the problem server itself? I have Retrospect client 6.1.130 running on my xServe.

The one thing that would hint that its not Retro’s fault is that in the afp logs, the CatSearch shows up for a specific connected afp user.

Just frustrating… my server slows almost once/day now and its not exactly easy to kick everyone out and restart afp.

[jpbuse]

[QUOTE][u]Quote by: pingu[/u][p]Sorry to hijack a little…

Can anybody give any hints on how the switch end should be set up for HP Procurve switches.

We have choices of ‘Trunk’ or ‘LACP’ in ‘HP speak’. As far as I can tell from the HP manual, LACP is proper 802.3ad and can be configured as either static or dynamic. Trunk is used for non-standard protocol trunks. HP recommend using Trunk when the host is using an unknown trunking protocol.

So it seems like LACP is best option, but Static or Dynamic???

Cheers
Dan[/p][/QUOTE]

Dan – You just need to enable LACP for the ports you will be connecting your machine to. That was I needed to do on my 4108gl switch. -Jason

[jpbuse]

FWIW

My setup is a xServe G5 DP 2.3Ghz, 10.4.8 PPC, 8GB RAM connected to xServe Raid running 1.5.1 firmware. My AFP server has crawled to a halt and continues to do so. I see the same entries on my afp access log as Moofo (CatSearch…)

My boot drive on the server is a software raid 1 of 2x 80GB drives.

Jason

[jpbuse]

Oops. I did read that actually, just didn’t sink in. I was hoping Universal would’ve fixed stuff magically but obviously not 🙁 Any word from Apple Ent yet on a fix/patch? I’ll most likely open a case on my own and reference this thread.

[jpbuse]

I’m glad I saw this thread… I’m seeing the exact same issues on my 10.4.8 xServe G5 dual 2.3Ghz server. I’m not sure if my users are doing a Finder style search but perhaps other programs are doing the searching as well. Things seemed to be working just fine until I moved all the core data from one sServe Raid to another. I’ve had these symtpoms before though and like you, a restart of the AFP server (if possible) fixes the slowness for a limited amount of time.

Is anyone running OS X Server Universal yet? Wondering if the AFP server in that build experiences all these fun issues.

[Author]

Posts