
Let there be light...
..at the end of the Lion tunnel
Attention Adobe! Read and learn: EMC fixed this during the Beta testing phase...
When the Lion roars, Celerra bleats and plays dead. El Reg has been told there is a problem with Mac OS X Lion clients accessing the Celerra filer array. The NAS head goes offline and failover doesn't kick in. An EMC patch is available and fixes the problem. Lioness with Celerra What am I supposed to do with this then? …
Look at all the manks on the iomega site, cos the nas dont work with Lion.
AFAIK Lion uses a new protocol, this was all published ahead of the release, Lion beta's were available, and the nas vendors haven't caught up, one reason appears to be a dispute with licensing of a piece of software required in the nas os to support the new protocol.
The filer head failover does work as intended, the probelm is that any standby heads will also panic if they also receive Lion CIFS connections and go into a reboot loop.
EMC did have a fix for this well in advance, however their usual advice to customers is to update the OS on these devices every 6 months or so. Any places that don't track the announcments closely and rely on the larger official release notes may have been caught out.
Normally, bugs in OS's don't take down the server they are talking to, just the system they're running on. The fact connecting a Lion client to a NAS head and taker the thing offline for everyone - and even worse, stop it from failing over to any reserve that may be available, that's a fairly significant and unusual thing.
You have misunderstood the issue. The issue is that a bug in the NAS's handling of the appletalk protocol that Lion uses causes the NAS to crash. The bug is clearly on the NAS, even if the bug is only exposed once exposed to Lion client.
Appletalk support is provided by the netatalk package, which on this NAS is netatalk 2.1.x, which supports AFP 3.2. Lion uses AFP 3.3, which netatalk 2.1.x gets wrong and crashes.
For Time Machine, your NAS needs to be running an AFP 3.3 compliant firmware - which means netatalk 2.2beta2 or later. This is from March, and few NAS makers have yet included it in their firmware particularly since it's still only at beta4.
For plain AFP shares, your NAS will still work fine.
For passworded AFP shares, the NAS either has to include the DHX2 library with netatalk (few do but it can often be added by a technically proficient user), or users must train their Lions to use deprecated security protocols instead, eg http://www.alexanderwilde.com/2011/04/os-x-lion-connection-error-with-afp-and-workaround/
For Windows shares, those will still work fine. Unless you're using a Celerra array without the latest patch, in which case your IT department is going to freak out massively.
I'm not suggesting that this isn't an EMC problem, but: How is this possible? Mac's CIFS connection is done with SAMBA, I don't notice any other implementations of SAMBA causing this behavior? What have Apple done to cause the problem, is it possible to make other SAMBA implementations behave in the same way?
Last week, Synology announced the latest version of their NAS OS (DiskStation Manager (DSM)):
http://www.synology.com/dsm/index_dsm3.2.php?lang=enu
They seemed particularly chuffed that they'd made its Time Machine server implementation Lion-compatible, so early - only catch is, this "new" DSM version is a beta, with all the caveats and gotchas that entails.
So, I now have to work out whether I'm willing to risk installing a non-downgradeable beta on our Synology box, to have the option of working Time Machine backups in Lion. Having said that, I'll probably wait for a while before upgrading to Lion anyway - though some new additions in that DSM beta (LDAP, CalDAV and syslog servers, virtual ISO disk mounter) look very tempting on their own...
Server fails due to client behaviour it can't handle. This is just plain awful - no excuses. The involvement of Lion is irrelevant. I have no idea whether the Lion client abides to whatever standards exist, and it doesn't even matter if it doesn't. Not talking to the client is one thing -- falling down on the job is just unforgivable.
OK - so unforgivable is a bit harsh. Bugs happen. I made a mistake once.
From EMC's face saving part, it comes down to when the problem was recognised and what steps were made to keep customers abreast of this.