FRS Randomly Loosing files.

  • Thread starter Thread starter James
  • Start date Start date
J

James

Hello I'm running DFS on two network attched storage boxes
(Win2k Server w/ SP4). They're exactly the same hardware and
communication between the two boxes is well established. These boxes
were set to host 3 diffrent shares. Archive, and Image collection, and
Spare area for quick file drops. The Images in the image directory get
pretty big, a couple of gigs at a time. For a while these machines
were ignored and the since they were using the default staging area
the FRS services would die on a regular basis.

I've since modifyed the staging are size to 4GB, since the largest
individual file never get past 3.5 GB I'm guessing this is an ok
ammount. I've since disable replication on all these shares and and
trying to slowy rebuild them in another shared directory. It seems
that FRS can't deal with large files and large numbers of files very
well.

I've adjusted the exclude filters so that "every thing" should be
replicated. I then moved a 7 gig directory into my testing area
"test2" Most but not all of the files were succuessfully replicated. I
ran win diff on the two shares and found that some intermitten random
collection of files were not replicated, these files had all vary
extensions and sizes. There was no common pattern between them.

My question is two fold.
1. What am I doing wrong? There seems to be no logic behind the
missing files. There are of varying extensions (.exe, dll, bz2,
etc...) and sizes. Is there some way to clean this up?

2. It seems that once frs has failed to replicate a file, it gives up
on that file for good. I've tried to tell frs to re-sync using the frs
diag utility. I've also gone as far as to restart the service so that
FRS retransmits the list of files to be replicated. Either the local
machine thinks these files have actually gone already, or the remote
machine is ingoring the requests. I can't seem to find a reason.


Any help provided would be greatly apperciated. Thanks

James
 
Hello James.

Are there any error messages in the FRS event log of any servers?

If a new file is placed on one server, does it get replicated to its
partners?

FRS is not suitable for large files that change frequently. At a basic
level, FRS is copying files over the network from one server to another. If
a 2 GB file gets updated every five minutes, that entire 2 GB file will
replicate every five minutes. This can obviously overwhelm the network and
cause other files that need to be replicated to be delayed. This may
explain why each replica is not in sync.

Best Regards,
David Fisher
Enterprise Platform Support
 
The even log reports no errors. The files them selves don't change
form weeks at a time. In either case I let it sit for a week. That
should be enough time to sync, considering our office was empty for
the week and there was no network usage (except for daemons like frs).
I fixed the staging area to small, not enough freespace issue over 2.5
weeks ago. It seems to give up on some files.
As I said previously, there is not common strain between files, they
all differ
in extension, size, mod date.

I did try making a dummy file in the scheme suggested by one of the
FRS diag docs. I made a file with the name of the partner on said
partner. It was replicated successfully. Granted this file had a zero
byte count. I then tried it on the other machine, still it sucessfully
replicates. Is there a way to "touch" files. So that FRS thinks, they
are new?
 
Back
Top