For all of you having services failing to start

root is by default enabled on OS X Server. It is the only account that can launch daemons running as another user which is mentioned in the plist file. Otherwise collabd won't run as _teamsserver which will generate other problems. And collabd was one of the first processes that would suffer from a failing postgres restart.

I'm quite sure the problem I've been facing is related to permissions but I've compared two systems bit by bit and can't find any differences.

The original system was installed with Lion Installer 10.7 and updated afterwards to 10.7.2. Maybe it is a library problem? I haven't compared them. It could be a timing issue as well as I sometimes had to reboot 4, 5, 6 times before postgres would even start again at boot time.

I've reinstalled the OS (Including server) using the 10.7.2 installer and this seems to have fixed the problem.

For the first time I've been able to issue the command serveradmin stop postgres followed by serveradmin start postgres without any problems. I did not wipe the partition but installed it over my boot volume. Copied all the data that was preserved in /Previous System back to assure the no mail, websites, agenda's etc were lost.

Also copied the old contents of /etc/ back. Will check the new /Library versus the saved one to see if I can spot anything.

launchctl works fine as well. I can load and unload services as root without problems.

The documentation of launchd is something I've read quite extensively. Unfortunate it doesn't provide me with any insights on how to troubleshoot it extensively.

Even though my system seems to be running fine now I've got to admit that I don't like to use the reinstall approach to solve problems if I don't know what I'm fixing.

Thank you sofar for your contributions.

Joe -

I took a look at this, and my root account was disabled. I enabled it, and have rebooted a number of times (including one full system crash) and there have been no hiccups with postgres.

It may be too early to tell, but I have never restarted this many times without problems.

Hmmmm. You may have stumbled onto something!

Today Apple confirmed that their is an issue with lauchd. Bug 10676168 describes the launchmsg() socket not connected problem. Launchd is used by many commands, including serveradmin, apachectl etc.

Let's hope that they will release an update soon.

Hi everyone,

Posting this in case it's useful to anyone.

I had a similar problem to that describe here: "error reading settings" when trying to look at the Wiki (and other) settings in Server.app.

My console log was full of:

1/18/12 2:26:51.418 PM org.postgresql.postgres: 2012-01-18 22:26:51 GMT FATAL: pre-existing shared memory block (key 0, ID 65536) is still in use

1/18/12 2:26:51.418 PM org.postgresql.postgres: 2012-01-18 22:26:51 GMT HINT: If you're sure there are no old server processes still running, remove the shared memory block or just delete the file "postmaster.pid".

1/18/12 2:26:51.418 PM com.apple.launchd: (org.postgresql.postgres[3602]) Exited with code: 1

1/18/12 2:26:51.418 PM com.apple.launchd: (org.postgresql.postgres) Throttling respawn: Will start in 10 seconds

So I followed the suggestion and did:

sudo rm -f /var/pgsql/postmaster.pid

That immediately appears to have addressed the problem, and it survives a reboot.

10.7.3 as been release since. I hoped it solved the problem, but the problem came back today. Are you guys having it too?

I'm sorry to say that I'm still having this problem too...

Yesterday I had to reboot 9 times to get the services started...

Whenever this problem shows its face I can not issue a sudo launchctl load /System/Library/LaunchDaemons/com.apple.collabd.plist, but I can unload it.

Trying to load it (after unloading it first) will display the launchctl_msg() socket not connected error.

Can you please verify if this is the case on your system as well?

The problem happens with any service. In my case, I created scripts that do several jobs automaticly for me. I tried unloading and reloading one and got the launch_mgs socket problem.

Just on a flyer, has this happened after you update to 10.7.3 AND doing the Terminal "dsenableroot" procedure.

When Brian tried that, prior to 10.7.3 it allowed many more reboots without a problem. I would think that doing that rather large 10.7.3 update might force the dsenableroot process to reset things again since it is obvious that a goodly part of 10.7.3 involved changes to the directory utility code.

I do not have Lion Server (it has been requested but so far no response from those who have to fund it) so all I can offer is this as a suggestion to try.

I just ran dsenableroot as.. root.. and rebooted. Let's see where it takes us 🙂

My next idea is to clear as single user /private/tmp and /private/var/tmp and recreate them.

Me too. Thanks for the reminder Joe. Best of luck!

Hello all,

I recently tried to set up OS-X Lion Server 10.7.3 on a mini mac server.

My (hopefully rather simple) goal is to implement the following services:

- DNS

- VPN

(-perhaps NTP)

OD

Calendar

AddressBook

FileSharing (for a start only group and user folders) with user data on server

I experienced the problems or symptoms You describe with 10.7.3. up to complete corruption

of the system. I have to clearly admit not being a professional in setting up those things.

Can You based upon Your experience suggest a setup or workaround that will possibly work?

I don't have a problem with starting with a fresh install, no vital data yet...

Further I don't mind working in the commandline or setting up some shell-scripts...

Kind regards and many thanks...

dsenableroot didn't solve the problem, it came back yesterday night

When I dug in and tried the dsenableroot after installing 10.7.3 I found that Apple changed the entire way enabling the Root account works. Now, you can enable Root from the Directory Utility repaated times BUT each time you try to enable it it now asks foir you to enter the Root password you will be using. This kind of adds Enable Root and Change Root password into one step. It no longer retains the Root password you had used previously.

All things considered, it looks like who ever tried to rebuild the Directory Utility for the Lion OS and its subsequent updates had better go back to school and this time they should stay awake in class so they can learn how to do it right! They still haven't figured out what to do to return the functionality that 10.6.8 and previous provided in this entire section of code.

I just went in single user mode. I did a fsck, everything was fine. I than emptied /tmp and /var/tmp, and rebooted. The machine kernel panicked on the reboot. The server is now reindexing via Spotlight (don't know why).

We will see in 3 or 4 days if it changed something...

It haven't changed a thing.

I tried reenabling "Reserve ressources for Server" and rebooted, problem came back 3 days later.