Server Lockup and Duplicate Clients in Admin

[Todd Wiener]

We have a big problem with our clients unable to connect to FM Server v11 after about an hour of use.

When this occurs, multiple iterations of the users appear in the Administrator panel. Sometimes these users can be forcibly disconnected, other times they cannot.

The only way we have found to remedy the problem is through a reboot of the server, which is not practical since we deploying in public areas for time limited applications.

Even then, the IPads are immediately seeking reconnection and the problem sometimes repeats almost immediately after the server reboot.

The problem seems to originate with iPad connections. The iPads only directly access one of two hosted files. (The second file contains data referenced by the first file.)

The screens of the iPads appear black (our splash layout color) with the rotating circle.

During periods of lockup, the server continues to run scripts, but users cannot connect to the databases.

We have FMS v11.0.2.217 running under OS X Sever 10.6.6 on a Mac Mini 2.66 GHz Intel Core 2 Duo with 8 Gigs of ram.

The clients include FM10 Pro, FM10 Pro Advanced, FM 11 Pro, FM11 Pro Advanced, and FM Go. All are the latest available versions.

Typical usage is 10 FM Pro and 15 - 25 FM Go, all connected wirelessly through Ruckus access points.

This seems to be a bug with FM Server, and a pernicious one at that.

Your ideas and suggestions would be most appreciated.

[lance123]

Hi Todd,

I have run into almost exactly the same problem minus the ipads. (Similar setup to yours) I posted on Fm forum and got the same solution from fm tech support as you have found. Restart the server. This is less than ideal.

I have previously found problems with ip v6 and have dropped that. I have also changed my network frame from 3 to 0 by doing the following sudo sysctl -w net.inet.tcp.delayed_ack=0 I am experimenting with this as it is meant to improve network performance. Please note the above change is me clutching at straws and I would not suggest you clutch at the same one's unless you have a development machine to experiment on. Which is what I am doing with these changes.

If you find a solution please post it here.

Regards,

Lance

[Steven H. Blackwell]

We have a big problem with our clients unable to connect to FM Server v11 after about an hour of use.

When this occurs, multiple iterations of the users appear in the Administrator panel. Sometimes these users can be forcibly disconnected, other times they cannot.

The only way we have found to remedy the problem is through a reboot of the server, which is not practical since we deploying in public areas for time limited applications.

Even then, the IPads are immediately seeking reconnection and the problem sometimes repeats almost immediately after the server reboot.

The problem seems to originate with iPad connections. The iPads only directly access one of two hosted files. (The second file contains data referenced by the first file.)

The screens of the iPads appear black (our splash layout color) with the rotating circle.

During periods of lockup, the server continues to run scripts, but users cannot connect to the databases.

We have FMS v11.0.2.217 running under OS X Sever 10.6.6 on a Mac Mini 2.66 GHz Intel Core 2 Duo with 8 Gigs of ram.

The clients include FM10 Pro, FM10 Pro Advanced, FM 11 Pro, FM11 Pro Advanced, and FM Go. All are the latest available versions.

Typical usage is 10 FM Pro and 15 - 25 FM Go, all connected wirelessly through Ruckus access points.

This seems to be a bug with FM Server, and a pernicious one at that.

Your ideas and suggestions would be most appreciated.

Poor choice of hardware in my view, but how much RAM is allocated to cache in the Console? It needs to be the full 800 MB for this number of users.

Steven

[lance123]

Hi Steven,

Thanks for that. I must say that there may be something to what you say. I have only had this problem on my development machine (core duo with 2 gig ram and 500 gig hard drive.) Which I know is vastly under-gunned for even that now. I have not had this on my Intel Mac Pro 8 gig ram 2.6ghz quad core. Which is the the production machine.

The Mac Pro has some minor issues with ghost license keys that we have to wait for a few minutes to time out if they have been forced to shut down unexpectedly. Otherwise no probs there.

Regards,

Lance

[Vaughan]

So can we take that a symptom of inadequate hardware is this problem with users not be disconnected?

[Todd Wiener]

Poor choice of hardware in my view, but how much RAM is allocated to cache in the Console? It needs to be the full 800 MB for this number of users.

Steven

Hi Steven -

The caches is at 256 MB currently, though I believe I've had it maxed out previously. I'll give it a try, but I don't see anything in the statistics to suggest it too low. Cache Hit % 75 to 100, Cache Unsaved % 0.

The Mac MIni server is well within FileMaker's specs and also we've had this problem with as few as 9 users (1 FM Pro on a laptop, and 8 iPads.) which suggests to me that this is more likely a bug or maybe file corruption.

The problem started occurring a few weeks ago. There were no hardware or software changes, just database tweaks. We also deployed for one show that went flawlessly. It's pretty vexing. Thoughts?

[Todd Wiener]

Hi Todd,

I have run into almost exactly the same problem minus the ipads. (Similar setup to yours) I posted on Fm forum and got the same solution from fm tech support as you have found. Restart the server. This is less than ideal.

I have previously found problems with ip v6 and have dropped that. I have also changed my network frame from 3 to 0 by doing the following sudo sysctl -w net.inet.tcp.delayed_ack=0 I am experimenting with this as it is meant to improve network performance. Please note the above change is me clutching at straws and I would not suggest you clutch at the same one's unless you have a development machine to experiment on. Which is what I am doing with these changes.

If you find a solution please post it here.

Regards,

Lance

Thanks Lance. It's good knowing that this is not an isolated incident. I hope FileMaker will address it, since server failures like these render the product unusable. We're looking into alternative platforms and in the meanwhile I'm doing a clean install of the server and reverting back to some previous versions of our database to see if file corruption might be at issue.

[NYPoke]

Ah, I found kindred spirits. I posted something myself, but this certianly seems to be what is happening to us.

In our case, it is worse. FM Server closes all the databases, after disconnecting everybody. Takes a long time to get 6 Gig open again.

With nothing else obvious, I had our Network Amin reboot. After reading this thread, turned out to be a good idea.

We have plenty of horsepower, memory & cpu. It is a virtual machine, but we've run without a problem for over a year. And then, on Friday the 13th, poof.

On the first shutdown, we got a message about a corrupt temporary file. On the second, we got no message at all, other than disconnects & files closing. Arrrgh.

[NYPoke]

Addendum. The one client, which seems to be in the middle of all this, cannot log back into FileMaker, after the server reboot. We have tried her on 2 machines & the Remote Desktop. Still, can't get her logged in. Everybody else & the server are all all fine.

Only thing I can associate with this is the Single Sign On/External Server Authentication. Well, her name is really long also. That isn't new though...

[Todd Wiener]

We have tested previous versions of our database and they do not display this behavior. Whatever is causing our problem seems to be triggered by file corruption that is not detected by FileMaker Server or FileMaker Pro. So, sadly, we are going to redo out database from the latest version that doesn't seem to have this problem. It does appear to be clearly a bug with Server. The same thing occasionally happens on other files connected with FM Pro Advanced. But it's much more rare.