Using both OSX 10.6 and 10.7 we've been encountering intermittent issues connecting to an AFP share. Sometimes if you disconnect and reconnect your network on the mac, the share will again be accessible. Other times it won't.
Right now for instance, I'm archiving a 500+ Gig dmg file to an AFP share. I went into finder to check on the current file size and it couldn't connect to the very same share I'm currently writing too! When I tried to click to connect-as button, it explained that another process was using the share and so it can't unmount it. WTF? So I can't connect, but I'm currently connected....and I can't unmount because I'm connected. Makes absolutely no sense. I could drop to a shell and go to /Volumes/Sharename and see the files that way, but could not open the share from finder.
So I waited a while....then tried again, and voila, its back accessible via Finder....presumably after some kind of timeout somewhere.
This is a brand new QNap 789-Pro with the latest firmware installed (3.7.2) and so I don't know if this is the behavior everyone has come to expect or not. I know from experience with FreeNAS that we did not have these sorts of issues before. I'm wondering if it might have something to do with the MTU size I specified on the TenGig interfaces (9000 bytes). We have a Cisco switch that fully supports jumbo frames, but I'm not sure about the Apple stack.
I've searched the forums here and found many mentions of AFP connection issues, but nothing that presents itself as intermittent and self-recovering. I've read up on the older compatibility issues with Lion and QNap when the DHX2 option is turned off, and ours is not. I've also read about issues caused by old hidden apple files on the share...and that its fixed when you remove those files, however we've taken no action and the AFP share eventually comes back.
I'm also not seeing incredibly good data transfer rates over gigabit ethernet with the QNap attached by full duplex dual-10gigabit SFP+ cables. It seems our recv link utilization rates are around 23 MB/s with 1 client writing over AFP. I think I've seen rates upwards of 100MB/s with an rsync inbound.
Does anyone else see similar issues? I'm most concerned about the AFP shares just refusing connections. The transfer rates can be lived with for now, but since Time Machine is running over AFP, I need AFP to be solid, and its simply not. I don't recall having this issue with FreeNAS, which likely uses the same underlying open-source software (afpd and avahi- although now that I check, it looks like freenas is running mDNSResponder). So either Avahi is whacked, or I've got something bad in my configuration.
thanks...
Byron
News