Jump to content
SNS Users' Forums
  • Announcements

    • Eric Newbauer

      Follow SNS on Twitter!

      Get notifications via Twitter! Looking for instant updates? We're now also announcing new versions and beta programs via Twitter. Follow Studio Network Solutions on Twitter. Thanks!

Recommended Posts

We've been struggling with GlobalSAN under OS X 10.7.5 server talking to a Dell Equallogic SAN and wanted to share for anyone else who will be experiencing problems.

If you are not aware, the Equallogic is an enterprise class SAN device. On one site we have two physical boxes in a single group providing over 150TB of iSCSI storage. Each device has multiple gigabit connections into our network backbone, and each gigabit link has it's own IP address.

We bought a license for GlobalSAN so that our OS X fileserver could utilise 10TB of storage for Mac clients using AFP.

Everything will work fine, until the Equallogic performs a load-balancing operation which basically disconnects a client causing it to reconnect, at which point it will be directed to the interface that is least busy.

When using GlobalSAN, this is effectively similar to someone yanking out the USB cable of an external USB drive, then quickly connecting it back in again. The problem is that from OSX's point of view, if the reconnection happens before the drive is dismounted, this results in a second volume with ' 1' appended to the end. i.e. if your iSCSI volume is called 'Data', you will end up after such an operation with a volume called 'Data 1' mounted, then 'Data' disappears.

The problem is this breaks Apple's file server which behaves as though the volume was simply unplugged.

*Sometimes* it recovers quickly enough to continue sharing files, however any sessions get disconnected. This is obviously extremely disruptive for all Mac clients who get disconnected from the fileserver and in some cases have to reboot. Other times it doesn't, which causes a denial of service to our users. We reboot the server in this case, but sometimes the server freezes on reboot and the crash report indicates the problem was the GlobalSAN extension. We are assuming that it basically has difficulty coping with the Equallogic load-balancing operation.

When it recovers, we repeatedly get the error in the system log:

kernel: GLO Warning: error during the recovery

The connection is also orange in the system preference.

We did contact GlobalSAN support (#208219) however a resolution was not forthcoming.

Due to the amount of complaints from our users we had to find an alternative solution, so we have purchased a license for the ATTO iSCSI client, and we have been running this now without a single issue for the last two weeks or so, so this would suggest a problem with the GlobalSAN client. It's a shame because this is literally the only issue we have had with the GlobalSAN software.

Hope this is of use to Equallogic users.

Share this post


Link to post
Share on other sites

Hi paulf (and other EQL users),

Case 208219 (October, 2014) led to engineering issue GLO-391, which is unfortunately at a dead end because we could not reproduce the reported behavior with our EqualLogic array, and the tcpdump requested to do further investigation was apparently not provided. (Apologies if we somehow missed it.)

If you have an EqualLogic array that is behaving similarly please comment here and/or open a case so that we can continue our analysis. We would love to have a tcpdump that captures this problem in action! To produce the dump:

Open Terminal and enter sudo tcpdump -i en0 -s 65535 -w ~/Desktop/tcpdump (where en0 is the iSCSI connection).

Reproduce the behavior, and then hit Ctrl+C in the Terminal to stop the tcpdump. The generated file will be on the desktop; please send it to our support team and reference this thread, or case 208219, or GLO-391.

Thanks,
Eric

Share this post


Link to post
Share on other sites

Hi paulf (and other EQL users),

Case 208219 (October, 2014) led to engineering issue GLO-391, which is unfortunately at a dead end because we could not reproduce the reported behavior with our EqualLogic array, and the tcpdump requested to do further investigation was apparently not provided. (Apologies if we somehow missed it.)

If you have an EqualLogic array that is behaving similarly please comment here and/or open a case so that we can continue our analysis. We would love to have a tcpdump that captures this problem in action! To produce the dump:

Open Terminal and enter sudo tcpdump -i en0 -s 65535 -w ~/Desktop/tcpdump (where en0 is the iSCSI connection).

Reproduce the behavior, and then hit Ctrl+C in the Terminal to stop the tcpdump. The generated file will be on the desktop; please send it to our support team and reference this thread, or case 208219, or GLO-391.

I supplied traffic captures of a period during the 'GLO Warning'; getting the precise moment that it fails however when it might involve a capture of terabytes of data transfer is not really feasible as it will happen at random a couple of times a week. We only notice it about once a week on a 'bad' recovery. This is a busy file server servicing the home folders of over 150 users of Adobe products to give you an idea of scale.

While I would like to help further, I cannot experiment with a live system under load - I'm sure you understand. The ATTO software continues to work perfectly however, so the available evidence does indicate a problem with GlobalSAN.

Share this post


Link to post
Share on other sites

×