10.6.3 frequent crashes

I am having similar problems. Upgraded to 10.6.3 yesterday and the server crashes (seemingly randomly) every few hours. I don't know what is causing the problem as I am a non-technical user.

Our computer is a late 09 mac mini server.

jpbuse:
Still seeing the problem? Anything useful in /var/log/system.log or other logs?

I think this crashing problem has existed for all versions of 10.6 server or it has for us. We have 18 Snow Leopard XServes running 10.6.3, I can not get any of them to work longer then 2 weeks with out crashing, some crash daily. I was hoping 10.6.2 would fix it and it did not same for 10.6.3. All servers are fresh installs, I even have 4 new Xserves with the OS installed from Apple and they all have the same problem. I have called Apple, had our SE out, had an Apple Integration specialist out and the best they could come up with was write a script to restart the servers every night. That does not even work because the crashes are so random. My current working theory is it has something to do with disk IO and Kerberos, if the servers are just running OD they do not crash as often but we need AFP and SMB so that is not an option. One thing I have found in the logs that no one I have talked to can tell me about is this. It shows up in the logs each time a user connects to a share.

Apr 21 15:15:29 server-safp-01 gssd[4510]: Minor error = 100006:
Apr 21 15:15:29 server-safp-01 gssd[4510]: Error returned by svc mach_gss_init_seccontext:
Apr 21 15:15:29 server-safp-01 gssd[4510]: Major error = 851968: Unspecified GSS failure. Minor code may

If anyone has any Idea how to make Snow Leopard Server more stable let me know

We have 4 xserves running, 1 is the OD master, the other 3 are OD replicas. They have all been randomly hanging. The hanging only started after updating them all to 10.6.3. Before the update to 10.6.3 they were running 10.5.8 in the same OD master/replica arrangement, all running very stable.
I did not upgrade the OS, I performed a clean (erase and install) installation. I did not import my 10.5.8 OD information, I setup a fresh clean OD system.

I have been leaving ssh sessions open on the servers and have been able to remain connected during the 'hang', however the output of most commands run during the hang is sporadic at least - is:

xserve1:/ root# reboot
-sh: /sbin/reboot: Permission denied
xserve1:/ root# tail /var/log/system.log
xserve1:/ root# ls -la /sbin/reboot
-r--r--r--@ 2433 root 264579 0 Apr 9 09:57 /sbin/reboot
xserve1:/ root#


When server is running ok:
xserve1:~ root# ls -la /sbin/reboot
-r-xr-xr-x 2 root wheel 55760 Feb 11 19:58 /sbin/reboot

As you can see, the uid mapping is completely screwed and the filesystem is reporting the reboot binary as being 0b in size! All commands above were run as root.

I have created a case with Apple support but they have reported no other reported issues with 10.6.3 server.
I have also disabled all 3rd party software running on these servers - radmind, networker, deploystudio. And even unbound the servers from active directory - as apple support informed me I would have to pay a fee to get my issue looked at since the machines were bound to AD.
Even after disabling all these 3rd party (AD plugin???) apps, the servers still freeze/hang.

I have tried rebooting them via LOM which has mostly no result, every now and them a LOM restart works but the majority of the time it does not.

Services currently running on the servers are:
AFP - some file sharing - NOT home directories.
NETBOOT
OPENDIRECTORY (kerberos disabled).

By the way, I have applied all current updates available from apple including the 10.6.3 v1.1 update.

Hey All,

I to am having the same issue with the server randomly hanging. The only fix is it just press the power button. The server is running OD/AFP/Netboot/DHCP/Nat/Firewall and only has 36 client machines and has 1200 user accounts but only a few login at one time. This is also running 10.6.3 on a brand new Xserve with SSD and 3TB Raid.

Your help would be fantastic Apple 🙂

Just a suggestion - but if you have Time Machine turned on it is known to cause crashes on 10.6.2 and 10.6.3 versions of Snow Leopard Server.

To find out more there are many threads on this topic in the Directory Services forum - look here for example: http://discussions.apple.com/thread.jspa?threadID=2404563&tstart=0

HTH

Hi, thanks for the suggestion, we do not run time machine on any of our servers.

Can I suggest that people having this issue that have current applecare agreements, contact apple and create a support case. The feedback I have had from Apple about this issue is that no one else is reporting any issues with 10.6.3 server.

I opened a case about this in November with Apple (Case ID 143620565), the problem has always existed in 10.6. I think it is starting to become apparent now that more people are moving to Snow Leopard Server. The problem has to do with file sharing lode, the more connections the faster it crashes. We have moved most of our shares off the 10.6 servers because of this problem and lack of a solution by Apple.

I have managed to get the output of dmesg during a system hang:
0 \[Level 3] \[ReadUID 0] \[Facility com.apple.system.fs] \[ErrType IO] \[ErrNo 6] \[IOType Read] \[PBlkNum 18901936] \[LBlkNum 0] \[FSLogMsgID 313908054] \[FSLogMsgOrder First]
0 \[Level 3] \[ReadUID 0] \[Facility com.apple.system.fs] \[DevNode /dev/disk0s2] \[MountPt > \[Path /private/var/log/asl/StoreData] \[FSLogMsgID 313908054] \[FSLogMsgOrder Last]
disk0s2: media is not present.

Considering disk0s2 is my system partition then I can see why this is causing issues!

After talking with some of the other admins that use server admin tools remotely on this server, it appears that the issue 'may' be related to using workgroup manager remotely. This particular admin had upgraded to 10.6.3 server admin tools using software update, but whenever adding a new plist for a group, it would cause this error (along with the following error in WGM "[MCXAdmin] Debug Message Unexpected error. File:/SourceCache/WorkgroupManagerPreferences/WorkgroupManagerPreferences-437/L owLevelEditor/PrefDetailsPI.mm Line:1542")

This dmesg error would appear on the OD master first, then trickle through to the OD replicas, rendering them all useless.

It's rather concerning that adding a plist to WGM would cause this sort of kernel level file I/O issue. We have since reinstalled the serveradmin tools on the 'suspect' desktop and are waiting for the system to fail again.

Message was edited by: Francois Herbert - Had to escape the square brackets in the dmesg log so they would appear

Same here.
New xserves, one OD server, one replica, one mail server often crash several times a day without any reason.
From what I see in system.log, the last entries before a crash are always from 'servermgrd'.
Over the last weekend I made sure that no 'Server Admin' application was running on any machine, including my admin workstation.
Guess what: No crash for three days.
Yesterday I had to make some changes using 'Server Admin' and the crashes happended again.

From my point of view, this issue is somehow related to 10.6.3.
I did not have any problems when working with 10.6.2 on the same machines.

Could someone confirm that this theorie might be plausible?

What I am seeing in my system log is:

May 4 10:49:33 ldap servermgrd[70]: Allocated size has grown to 3M. Number of allocations is -4980. Exiting to clear possible memory leak.
May 4 10:49:33 ldap com.apple.launchd[1] (com.apple.servermgrd[70]): Exited with exit code: 12
May 4 10:49:33 ldap servermgrd[21158]: servermgr_accounts: got error 5300 trying to auth to local LDAP node