Finder crashes after xsan finder copy, Apple is stumped

jsiegers · Could work for Apple Joined: 21 Oct 2008 Posts: 41

Hi Moe,

We're not even able to log out.
Do you also have a large extent count?

The problem is with Xsan software for the client. We've been talking to Apple about this bug. But I don't know if it's fixed in Xsan 2.3

Jasper

Moe · Xsan Master Joined: 04 May 2006 Posts: 61

Hi jsiegers

jsiegers · Could work for Apple Joined: 21 Oct 2008 Posts: 41

When a file is heavily fragmented the machine requires more memory. We've noticed that on 10.5 only FCP is affected but sometimes when you wait long enough (15 minutes) the project is loaded like nothing ever happened.

I know (but haven't seen it myself) that on 10.6 the Finder should be affected as well.

You could simply check the extent count even on StorNext systems by using the snfsdefrag -c command like this:

wrstuden · Xsan Master Joined: 04 Jan 2008 Posts: 99

jsiegers · Could work for Apple Joined: 21 Oct 2008 Posts: 41

And there was much rejoicing: Apple has released Xsan 2.2.2 which resolves all our issues!

http://support.apple.com/kb/HT4804

mw10dot1 · Xsan Master Joined: 06 Dec 2006 Posts: 106

Hi jsiegers

Where you running 10.6.8? As the posts from apple have two different requirements. One says 10.6 the other says 10.6.8

thanks

Michael

jsiegers · Could work for Apple Joined: 21 Oct 2008 Posts: 41

We have yet to install this update.
Have to plan some downtime first.

Apple said to me that you'd have to have 10.6.8 for this update.
We're now running 10.6.7 on the servers and 10.5.8 on our clients.
We can now finally upgrade our clients to 10.6.8 now this is fixed.

tomwelch · Been around the blocks Joined: 30 Sep 2009 Posts: 25

Hi all.

I think I can say that this is affecting me too!

Where it gets really annoying for me is that we re-share our Xsan volume from two APF servers. These are the machines that gets the spinning wheel and 'empty' folders. Of course this has a knock on affect to the 50-70 AFP connections on any given day. The afp process gets stuck and all connections get a spinning beach ball in their finders. Sometimes the AFP servers will kernal panic, sometimes I have to force them off.

I've been running jsiegers' script (great job by the way) and we have number of files with high extends.

Right now we're getting this a few times a week! So I'm considering running the 2.2.2 update this weekend. We only have 2 MDCs and 3 clients (all Xserves). Everything is running OS X Server 10.6.7 and Xsan 2.2.1.

If I backup my servers system volumes before updating to 10.6.8 + Xsan 2.2.2, and it all goes wrong, I can just restore my backups right? Anybody know if the 2.2.2 update alters the Xsan file system content? Obviously, I'll have a full backup!!!

We've had this problem for quite some time (it's been progressively getting me frequent) so heres hoping this update fixes it!

Finally, has anyone already installed this update?

Thanks for all the posts! I'm now (slightly) happier that I know whats going on, and that it's not just me.

Tom

jsiegers · Could work for Apple Joined: 21 Oct 2008 Posts: 41

Hi Tom,

I'd suggest that you defrag your system first. That way you don't have the spinning wheel for now.
Than make images of your servers. That way you can easily go back.

I've been testing a preview build of this version and I've found it very smooth and stable. But please test it first if you can.

tomwelch · Been around the blocks Joined: 30 Sep 2009 Posts: 25

Hi jsiegers

Thanks for replying!

Unfortunately we don't have a Xsan test environment. I usually hold off for a few months to see what others report about new updates but in this case I'm not sure we can.

Am I right in thinking that repeated defragmentation is not recommended on an Xsan volume? I'm tempted to run your script say once a week? As it only defrags individual files do you think that will be OK?

Thanks.

t.

abstractrude · Xsan Master Joined: 13 Mar 2008 Posts: 863

I had a customer that i fought with for months about this once. eventually they listened to me and didnt fill their volume up, problem went away.

tomwelch · Been around the blocks Joined: 30 Sep 2009 Posts: 25

abstractrude · Xsan Master Joined: 13 Mar 2008 Posts: 863

yes but at once point it wasnt. and eventually freespace fragmentation affects your entire filesystem. my only point is. i have SANS that have a 2 year uptime and if you check a file for extents, it has 1. i would pull all your data off and rebuild your volume for good measure. if thats not possible constantly running defrag will only make things worse.

tomwelch · Been around the blocks Joined: 30 Sep 2009 Posts: 25

Thanks abstractrude. I actually rebuilt the volume from scratch about three months ago and added more storage. This is the first time the volume has got down to 20% free space. With the backup volume it's not too hard to redo though so maybe I'll do it again. What kind of free space do you recomend?

tomwelch · Been around the blocks Joined: 30 Sep 2009 Posts: 25

So, over the weekend I didn't get to run the 2.2.2 update for various reasons. I did however, mainly our of curiosity, run jsiegers's script on our backup SAN volume. This volume is cloned using PresSTORE. It is identical in setup and hardware to the live volume. There is no where near as much fragmentation on the backup volume and that got me thinking.

Next I preformed some tests putting files on the SAN via the AFP server and directly. Anything added to the volume via an AFP share ends up being fragmented, anything added directly to the volume is fine.

Has anyone else seen this with re-sharing?

Thanks in advance,

Tom