Looking for some advice.
Background:
We have a Dell PE R200 with 2008SP2 that suffered heat damage.
We have one new Dell R530 running 2012r2 and an IBM System X3630M3 running 2012. The IBM was inherited from another division, and reconfigured to our needs.
Originally there was no DFS setup until we got new servers. We decided with some new servers that we would use DFS to improve data availability. The original plan was to pre-seed the new servers from backup tapes, use robocopy to correct
file/folder permissions on the new servers using the old servers as the source. Then set DFS to target all four servers with the R200 and an older Dell 2950 (server 2003sp3) as the primary sources to the newer servers to synch it all. This worked great with out our 2950 shares and user home folders which totaled about 1TB between it the sources. The problems occurred with the 8TB of static archived data which held of on incorporating until
the dynamic data was verified good and operational. Then we added in this static data the same way as we did the dynamic.
Five days into add the static sources I noticed the IBM is still waiting on initial replication with 69,139 files in the backlogged sending and 69369 in backlogged receiving. The IBM also has a backlog receiving transaction of 139K for the
second set of static data. I don't see any backlog on sending for this dataset on any of the servers.
I did some file hash testing last week with the “DFSR FileHash Comparison Pre-Stage” script, which I found in the gallery, and it indicated that IBM matched the R200 but that the R530 was not matching either of them and I started noticing
performance issues with IBM; disk performance was maxed which was effecting the print server and other DFS namespaces on this server. We ended up disabling the IBM associated DFSR groups for these 2 sets of
static data.
So this week I tried diving into what is the issues were with static data, but the file hash script I used the week before will not work. Not sure what I am doing different this week versus last week with the script.
After looking at the DFS reports I ran last week I also see that the R530 has 96,399 backlogged sending transactions for one set of data but the other set is good and neither set had any receiving on this server. Which struck me as odd since
the file hashes did not match last week.
I was under the impression that pre-seeding from backups and then using robocopy to set file/share permissions to match the original on the R200 would be the quickest way move the data, initialize and get synched. At the rate things are going it will be months before my data is properly synched and not causing disk latency issues with the IBM. The R200 has disk performance and CPU performance issues due to a overheat condition that occurred earlier this year which only compounds the synching.
Question:
So how should we go about setting up DFS on the IBM with minimum impact on the server performance and the network and complete it in a timely manner.
I am, thinking we might need to try DFS cloning on the IBM, but that would require us to tear down all the DFS namespaces associated with the IBM and start over with only the R530 set up. As I understand it the DFS cloning can only be done with Server 2012
or higher. But how should I pre-stage the data if I go down this road? And of course I would have to shutdown the DFS on the R200 and the 2950, which is not a huge issue as we plan to retire both of those, but I whats the safe way to
remove those 2 from DFS. When I was testing DFS awhile back I mucked up the removal of the test setup and had to do some manual cleanup plus I discovered it deleted the test data. I believe the correct way now is to disable the associated
server replication groups, then delete the associated target folder then delete the server from the name space, correct?
I will probably tear down the R530 static data namespaces as well and then restore the data from tape, set the file and folder permission then rebuild the namespace for this data. Then DFS clone the R530 server to the IBM? How should I pre-stage
the data this time on the IBM?
Some addition server info:
The IBM – Xeon E5630, uses RAID 1 for the OS and RAID 5 for the data, RAID 5 is using 4x 4TB SAS HDD and the controller is a IBM M5014 (which is a LSI (3000?) card) and has 24GB RAM, 2x1GB NICs using Link Aggregation (L3_L4 LACP).
The R530 - dual socket Xeon E5-2640v3, uses RAID 1 for OS and RAID 5 for data with 5x 4TB HDD using a Dell Perc 730 (the 2GB ver.) and 128GB RAM. The R530 has 2x10GB Converged NIC and 4 1GB NICS. I have the R530 connected using the 10GB the plan is to use the 1GB for Hyper-V down the road.
The static data in question currently resides on 2 Dell PV MD1000s connected to the R200 using a perc6 external Raid controller with the InfiniBand connectors. Once we get this archived data moved and properly setup we plan to re-configure the MD1000 using custom SAS cables with the IBM and R530 using 12GB HBA’s.