Mac OSX desktop dropping connection with multiple copy processes & large files

Question

The servers are 6.5 SP3 running NFAP, the MAC OSX is 10.4.2 updated. The
volume the macs are using is part of a cluster. The users mount the volumes
on their macs and everying is for the most part fine. If they grab a bunch
of files and copy them from desktop to server it's fine as long as it's only
a single copy process. The users are part of the hi-res department and the
files can be 1GB or larger. If they drag one or more large files, and then
while that's copying they drag some more files, so both copy processes are
running at once....quite often the volume will dismount from the desktop and
you will get unable to copy because some resource is unavailable. Sometimes
the finder crashes, sometimes not. Often the files that were partially
copied get locked and the users needs to reboot their Mac in order to delete
them. I'm getting pretty desperate hear, anyone have an idea what's going
on. I don't know if this is a Tiger thing or a large file thing or a
multiple copy stream thing, a netware thing or a mac thing.....we have
hundreds of other users running OSX 10.3 and earlier who are not reporting
this problem, but they also don't copy files that size. Someone please tell
me they have seen this before....thanks very much. Oh, before going to 6.5
and NFAP the servers were 5.1 with Prosoft server and they never had the
problem.
Jake

Answer 1

Thanks for your help, I have incidents open now with Apple and Novell, I
hope one of them can provide something for us. We tried applying 6.5 SP4 to
a test server....the problem still happened but was "better", the copy
operations still quit but with SP4 applied the volume did not dismount....or
if it did it remounted automatically because it was still connected after
OKing through the copy errors.
"Jeffrey D Sessler" <[email protected]> wrote in message
news:[email protected]...
>I tried two 2GB files. No problems at all but I'm in a 100% end-to-end
>Gigabit environment. My server storage is also a very-fast SAN.
>
> Best,
> Jeff
>
>
> "Jacob Shorr" <[email protected]> wrote in message
> news:[email protected]...
>> Jeffrey,
>>
>> Have you tried the exact same test, dragging say two 500MB files in
>> seperate
>> copy operations? I hear what you're saying about the 10/100 link, but we
>> don't run gigabit to the desktops, and we're not going to anytime soon.
>> Even if that could resolve the issue we need something kind of other fix
>> for
>> our infrastructure. I will look into any errors on the switch.
>>
>> "Jeffrey D Sessler" <[email protected]> wrote in message
>> news:[email protected]...
>>> Well, considering that I'm not seeing the issue on my 10.4.2 machines
>>> against my 6.5Sp3 servers, I'm not sure what you should do at this
>>> point.
>>> Since you say that the 10.3 machines don't have an issue, it makes it
>> sound
>>> to me like this is an Apple issue.
>>>
>>> The logs point at a communication issue... Is there anyway to get that
>>> Mac
>>> on to a Gigabit connection to see if you can duplicate it?
>>>
>>> The other option is to wait for 10.4.3 to be released and see if the
>> problem
>>> goes away.
>>>
>>> Again, on only a 10/100 link, one copy of a large file _will_ saturate
>>> the
>>> link.Perhaps 10.4.2 has an issue with this?
>>>
>>> Also, when you're doing the copy, what to the error counters in the
>> switches
>>> say?
>>>
>>> Jeff
>>>
>>> "Jacob Shorr" <[email protected]> wrote in message
>>> news:[email protected]...
>>> > There are definately no mis-matches. This has been checked and
>> re-checked
>>> > a
>>> > dozen times. It's only on 10.4......we can replicate it on every 10.4
>>> > machine, and we cannot replicate it on any machine that is 10.3. What
>>> > should I do to go about getting this fixed, should I be contacting
>>> > Apple
>>> > or
>>> > Novell? The speed is always good until it actually decides to drop
>>> > and
>>> > cut
>>> > off.
>>> >
>>> >
>>> > "Jeffrey D Sessler" <[email protected]> wrote in message
>>> > news:7jj%[email protected]...
>>> >> Looks like communication between the Mac and the Netware server is
>>> > dropping.
>>> >> AFP in 10.3 and 10.4 support auto-reconnection but I'm sure that it
>> will
>>> >> fail the copy process.
>>> >>
>>> >> I'd first check to make sure that there are not any mis-matches on
>>> >> the
>>> >> switch e.g. the Mac is set to Auto (as it should be) but someone has
>> set
>>> > the
>>> >> switch to a forced mode. Both should be auto. A duplex miss-match
>>> >> could
>>> >> cause the Mac not to see the heart beat back from the Novell server.
>>> >>
>>> >> Like I said, if the workstation is only on 10/100, a single copy
>> process
>>> > on
>>> >> a G5 Mac will saturate that link. Adding more concurrent copies will
>> only
>>> >> result in everything slowing down and taking longer, or you'll get
>>> >> the
>>> >> dropped connections.
>>> >>
>>> >> Best,
>>> >> Jeff
>>> >>
>>> >>
>>> >> "Jacob Shorr" <[email protected]> wrote in message
>>> >> news:Ybc%[email protected]...
>>> >> > Take a look at the last entries in the system log right after it
>>> > happened,
>>> >> > let me know if it means anything to you. Thanks.
>>> >> >
>>> >> > Sep 29 13:26:10 yapostolides kernel[0]: AFP_VFS afpfs_mount:
>>> >> > /Volumes/FP04SYS11, pid 210
>>> >> > Sep 29 13:31:13 yapostolides kernel[0]: AFP_VFS afpfs_Reconnect:
>> doing
>>> >> > reconnect on /Volumes/FP04SYS11
>>> >> > Sep 29 13:31:13 yapostolides kernel[0]: AFP_VFS afpfs_Reconnect:
>>> > connect
>>> >> > to
>>> >> > the server /Volumes/FP04SYS11
>>> >> > Sep 29 13:31:13 yapostolides kernel[0]: AFP_VFS afpfs_Reconnect:
>>> > Opening
>>> >> > session /Volumes/FP04SYS11
>>> >> > Sep 29 13:31:13 yapostolides kernel[0]: AFP_VFS afpfs_Reconnect:
>>> > Logging
>>> >> > in
>>> >> > with uam 2 /Volumes/FP04SYS11
>>> >> > Sep 29 13:31:13 yapostolides kernel[0]: AFP_VFS afpfs_Reconnect:
>>> >> > Restoring
>>> >> > session /Volumes/FP04SYS11
>>> >> > Sep 29 13:31:13 yapostolides kernel[0]: AFP_VFS
>>> >> > afpfs_MountAFPVolume:
>>> >> > GetVolParms failed 0x16
>>> >> > Sep 29 13:31:13 yapostolides kernel[0]: AFP_VFS afpfs_Reconnect:
>>> >> > afpfs_MountAFPVolume failed 22 /Volumes/FP04SYS11
>>> >> > Sep 29 13:31:13 yapostolides KernelEventAgent[43]: tid 00000000
>>> >> > received
>>> >> > VQ_DEAD event (32)
>>> >> > Sep 29 13:31:13 yapostolides kernel[0]: AFP_VFS afpfs_Reconnect:
>>> > posting
>>> >> > to
>>> >> > KEA to unmount /Volumes/FP04SYS11
>>> >> > Sep 29 13:31:13 yapostolides KernelEventAgent[43]: tid 00000000
>>> >> > type
>>> >> > 'afpfs', mounted on '/Volumes/FP04SYS11', from
>>> >> > 'afp_0TQCV10QsPgy0TShVK000000-4340.2c000006', dead
>>> >> > Sep 29 13:31:13 yapostolides KernelEventAgent[43]: tid 00000000
>>> >> > found
>> 1
>>> >> > filesystem(s) with problem(s)
>>> >> > Sep 29 13:31:13 yapostolides kernel[0]: AFP_VFS afpfs_unmount:
>>> >> > /Volumes/FP04SYS11, flags 524288, pid 43
>>> >> >
>>> >> >
>>> >> >
>>> >> >
>>> >> > "Jeffrey D Sessler" <[email protected]> wrote in message
>>> >> > news:GH%[email protected]...
>>> >> >> We move large files all the time under SP3 with no issues however,
>>> > there
>>> >> > are
>>> >> >> several finder/copy/afp issues in Tiger that are do to be fixed in
>>> >> >> 10.4.3.
>>> >> >>
>>> >> >> Also, if you have any type of network issue such as duplex
>> mis-matches
>>> > or
>>> >> >> are running say, only a 10/100 network, a single Mac can not only
>>> >> >> transfer
>>> >> >> more than 10MB/sec (filling the network pipe) or generate so many
>>> >> > collisions
>>> >> >> (duplex mis-match) that you could drop communication to the
>>> >> >> server.
>>> >> >>
>>> >> >> What type of server (speed, disks, raid level, NIC speed) and what
>>> >> >> type
>>> >> >> of
>>> >> >> network (switched gigabit, switched 10/100, shared 10/100, etc.)
>>> >> >>
>>> >> >> How long does it take to copy that single 1GB file to the server?
>>> >> >>
>>> >> >> Does a single copy process always work?
>>> >> >>
>>> >> >> Jeff
>>> >> >>
>>> >> >> "Jacob Shorr" <[email protected]> wrote in message
>>> >> >> news:[email protected]...
>>> >> >> > The servers are 6.5 SP3 running NFAP, the MAC OSX is 10.4.2
>> updated.
>>> >> > The
>>> >> >> > volume the macs are using is part of a cluster. The users mount
>> the
>>> >> >> > volumes
>>> >> >> > on their macs and everying is for the most part fine. If they
>> grab
>>> >> >> > a
>>> >> >> > bunch
>>> >> >> > of files and copy them from desktop to server it's fine as long
>>> >> >> > as
>>> > it's
>>> >> >> > only
>>> >> >> > a single copy process. The users are part of the hi-res
>> department
>>> > and
>>> >> >> > the
>>> >> >> > files can be 1GB or larger. If they drag one or more large
>>> >> >> > files,
>>> > and
>>> >> >> > then
>>> >> >> > while that's copying they drag some more files, so both copy
>>> > processes
>>> >> > are
>>> >> >> > running at once....quite often the volume will dismount from the
>>> >> >> > desktop
>>> >> >> > and
>>> >> >> > you will get unable to copy because some resource is
>>> >> >> > unavailable.
>>> >> >> > Sometimes
>>> >> >> > the finder crashes, sometimes not. Often the files that were
>>> > partially
>>> >> >> > copied get locked and the users needs to reboot their Mac in
>>> >> >> > order
>>> >> >> > to
>>> >> >> > delete
>>> >> >> > them. I'm getting pretty desperate hear, anyone have an idea
>> what's
>>> >> > going
>>> >> >> > on. I don't know if this is a Tiger thing or a large file thing
>> or
>>> >> >> > a
>>> >> >> > multiple copy stream thing, a netware thing or a mac
>>> >> >> > thing.....we
>>> > have
>>> >> >> > hundreds of other users running OSX 10.3 and earlier who are not
>>> >> > reporting
>>> >> >> > this problem, but they also don't copy files that size. Someone
>>> > please
>>> >> >> > tell
>>> >> >> > me they have seen this before....thanks very much. Oh, before
>> going
>>> > to
>>> >> >> > 6.5
>>> >> >> > and NFAP the servers were 5.1 with Prosoft server and they never
>> had
>>> >> >> > the
>>> >> >> > problem.
>>> >> >> >
>>> >> >> > Jake
>>> >> >> >
>>> >> >> >
>>> >> >>
>>> >> >>
>>> >> >
>>> >> >
>>> >>
>>> >>
>>> >
>>> >
>>>
>>>
>>
>>
>
>