ONTAP Discussions

Backuping up CIFS Shares

JRAYKOWSKI
18,716 Views

Hello,

  We brought a new FAS-2240 online starting in December and I have been working to move it into production and so far so good.  I have FC LUNs and CIFS Shares up and running in a test production environment and it is working as advertised.  The next hurdle is backups.  I have been reading over come of the posts here and it seems that unless we use NDMP to backup data the only other option would be to present the CIFS Shares as iSCSI LUNs as well to the backup server as new disks and conduct backups that way.

  Are there any issues with presenting active CIFS shares as iSCSI LUNS as well?  In most situation all iSCSI activity will take place after roduction hours and it is just being used to copy data for backups purposes.

TIA

Jim

1 ACCEPTED SOLUTION

adrianlowe
18,700 Views

Hello Jim,

You might find the following post useful regarding different NDMP backup options:

FAS2220+ Entry Level FAS backup to Tape - Connectivity options (SAS and NDMP) (https://communities.netapp.com/thread/25746)

I hope this is of some use.

Adrian

p.s. CIFS/NFS is file level, iSCSI/FC is block level, the flexvolumes have different properties set for very different reasons. NetApp has great features for providing more efficiency backups using the snapshots & snapvault for D2D backups and then it will allow you to backup to tape more effectively so you don't have to backup to tape in-efficiently on a daily base.

View solution in original post

19 REPLIES 19

aborzenkov
18,645 Views

There is no way to present CIFS shares as iSCSI LUN. Snapshots provide quite efficient backup for file shares.

Отправлено с iPhone

09.01.2013, в 20:28, "Jim Raykowski" <xdl-communities@communities.netapp.com<mailto:xdl-communities@communities.netapp.com>> написал(а):

<https://communities.netapp.com/index.jspa>

Backuping up CIFS Shares

created by Jim Raykowski<https://communities.netapp.com/people/JRAYKOWSKI> in Products & Solutions - View the full discussion<https://communities.netapp.com/message/97915#97915>

Hello,

We brought a new FAS-2240 online starting in December and I have been working to move it into production and so far so good. I have FC LUNs and CIFS Shares up and running in a test production environment and it is working as advertised. The next hurdle is backups. I have been reading over come of the posts here and it seems that unless we use NDMP to backup data the only other option would be to present the CIFS Shares as iSCSI LUNs as well to the backup server as new disks and conduct backups that way.

Are there any issues with presenting active CIFS shares as iSCSI LUNS as well? In most situation all iSCSI activity will take place after roduction hours and it is just being used to copy data for backups purposes.

TIA

Jim

Reply to this message by replying to this email -or- go to the message on NetApp Community<https://communities.netapp.com/message/97915#97915>

Start a new discussion in Products & Solutions by email<mailto:discussions-community-products_and_solutions@communities.netapp.com> or at NetApp Community<https://communities.netapp.com/choose-container.jspa?contentType=1&containerType=14&container=2068>

JRAYKOWSKI
18,645 Views

  I have a requirement to tape out everyday so I need to backup these shares.  Is there another way besides using NDMP?

TIA,

Jim

aborzenkov
18,646 Views

The only other way is to map share and backup it from client. Also, for NDMP you do not always need tape, attached to filer. Most backup software I know supports three way NDMP backup, when data stream is sent to media server (tape server) via LAN.

Отправлено с iPhone

09.01.2013, в 21:20, "Jim Raykowski" <xdl-communities@communities.netapp.com<mailto:xdl-communities@communities.netapp.com>> написал(а):

<https://communities.netapp.com/index.jspa>

Re: Backuping up CIFS Shares

created by Jim Raykowski<https://communities.netapp.com/people/JRAYKOWSKI> in Products & Solutions - View the full discussion<https://communities.netapp.com/message/97917#97917>

I have a requirement to tape out everyday so I need to backup these shares. Is there another way besides using NDMP?

TIA,

Jim

Reply to this message by replying to this email -or- go to the message on NetApp Community<https://communities.netapp.com/message/97917#97917>

Start a new discussion in Products & Solutions by email<mailto:discussions-community-products_and_solutions@communities.netapp.com> or at NetApp Community<https://communities.netapp.com/choose-container.jspa?contentType=1&containerType=14&container=2068>

adrianlowe
18,701 Views

Hello Jim,

You might find the following post useful regarding different NDMP backup options:

FAS2220+ Entry Level FAS backup to Tape - Connectivity options (SAS and NDMP) (https://communities.netapp.com/thread/25746)

I hope this is of some use.

Adrian

p.s. CIFS/NFS is file level, iSCSI/FC is block level, the flexvolumes have different properties set for very different reasons. NetApp has great features for providing more efficiency backups using the snapshots & snapvault for D2D backups and then it will allow you to backup to tape more effectively so you don't have to backup to tape in-efficiently on a daily base.

JRAYKOWSKI
18,645 Views

Adrian,

  I cannot seem to follow the link.  I asks me to register...um I already thought I did that which is why I am here.  Can you give me the quick down and dirty please.

TIA,

Jim

adrianlowe
18,645 Views

Indeed you should be able to access it as its part of the same community. Please see the attachment. It's not official documentation but hopefully it helps.

Heres the link again, just in case https://communities.netapp.com/thread/25746 you may have to login again!?

Let me know your thoughts on the doc.

JRAYKOWSKI
18,644 Views

Adrian,

  Woo Hoo it works.  Thanks for all your help.  I added our FAS-2240 as a NDMp Device in Backup Exec and am running a test backup now.  Kind of slow, 600MB/min, but it is working.

Thanks,

Jim

adrianlowe
18,644 Views

Excellent just out of curiosity what set up did you use? from what you stated above I'm guessing it ethernet connected to your backup exec server and therefore I'd probably say 600MB from a 1gb port isn't too bad although you have raised a very interesting topic which I wouldn't mind understand more on.... Backup transfer rates

if directly connected via FC then I would expect up to the 1, 2 or 4gb throughput. If its going over the ethernet via the ndmp server then your initially capped at the 1gb port HOWEVER this may lead on to discussing whether you can have VIF's / IFgroups which can help increase that throughput, if not for throughput then for resiliency.

You may want to try a few different options. Maybe having a NetApp VIF/IFGroup set to either multiple mode or LACP mode (depending on whether your switch or server support this) this will then give you two bonded ports however you will only initially have 1 IP address assigned across the two ports and therefore you are still only capped at 1gb speed for each session (IP pairing session) connectivity. You can then add an additional IP alias to this VIF/IFGroup which should then allow an additional 1gb session which should then increase the throughput (ideally). This would also depend on the sessions transfering and whether its just one large session. you then also have to ensure the server is also exceeding the 1gb throughput capacity if its connecting to a switch and then from the server to the Tape device.

What connectivity do you have from your server to your tape device (FC or SCSI or other)?

Thanks in advance,

Adrian

JRAYKOWSKI
18,644 Views

Adrian,

  Currently backup exec is running on a HP DL-380G5 with two HP NICs Teamed.  The tape unit  is a MSL-6000 with dual drives connected via SCSI.  Currently our biggest jobs come from a NAS Cluster and I average about 1200MB/min out of them with single NICs.  They do however have the BE Agent installed so there is probably some good compression.

  I will let this test run and see how it goes.  If I cannot get the performance I need I will creat a new VIF using two NICs and see what performance I can get out of that.  I wil probably start with multiple mode and if that does not get any better performance I will go to LACP.  Having performed some initial test with both modes I seem to get better performance out of multiple mode so far.

Jim

adrianlowe
12,886 Views

Hello Jim,

Excellent. Don't forget to try adding the IP alias's (typical assign an IP Alias for the number of ports within that VIF, i.e. if you had 4x1GB ports in a VIF, apply 4 IP addresses to that vif) to the VIF'ed ports too.

Let me know what results you get for each method.

Thanks for all your responses. Have a great weekend.

Adrian

p.s. I'm assuming your running all this on ONTAP 8.1.2?

JRAYKOWSKI
12,887 Views

Adrian,

  We are running ONTAP 8.1.1 for now and will upgrade soon.  Initial results were not so good with one NIC assigned to NDMP Backups, 494.9GB, averaged 668MB/min, and it took 24hr and 2 min.  Next to double and see what happens.  Will run next test next week have weekend full to get done first.  I will keep you informed.

Jim

JRAYKOWSKI
12,887 Views

Adrian,

  Wanted to provide an update. on PCM-1, filer1, I created a VIF with e0c and e0d.  I assinged it one IP and the name filer-1-backup.  I added it as a NDMP Device in Backup Exec,we are running version 2010 R3, and am running another test backup n the same volume I did last week.  So far after about 15 monites we are running at 2180MB/min and have backuped up 37.8GB of data.  So a huge improvement.

  Will keep you informed.

Jim

JRAYKOWSKI
12,886 Views

Adrian,

  Ned to correct yesterday and provide some additional information.  So yesterday I created a new VIF assigning e0c and e0d but I forgot to run ndmpd.preferred_interface and change it to vif3.  So when I ran the test it was actually assigned to e0d.  Throughput was wonderfull as I said but I wanted to see what happens when I assigned it to vif3.  So I made the chnage and throughput was terrible, in 7 hours it only backed up 11,805,696 bytes.  Arg!  So I chaned it back to e0d and again throughput was wanderfull.

Question is why?

TIA,

Jim

adrianlowe
12,886 Views

Hello Jim,

Thanks for the feedback so far.

Just to clarify a few things....

FAS2240 ONTAP 8.1.1

Backup Exec 2010 R3

NDMP backup method using ethernet connectivity from NetApp to servers. Using SCSI from Server to Tape.

NetApp configs

Test 1 (initial config). using 1 phyiscal NIC (with no VIF) assigned to NDMP backup. Transfered 494.9GB averaging 668MB/s - duration 24hrs & 2min to complete.

Test 2. Using a VIF (single, multi-mode or LACP?) pairing e0c+e0d using 1 IP address (with NO ndmpd.preferred_interface configured) assigned as an NDMP device to Backup. Transfered 37.8GB averaging 2180MB/min - duration 15mins. Note this transfer was going via e0d (although it was still in a VIF. If it was via a VIF then it would be the same set up as test 1?)

Test 3. Using same settings as Test 2 but now WITH ndmpd.preferred_interface configured to the VIF3 (which consists of e0c+e0d). Transfered ONLY 11,805,696 bytes in 7 hours.

I guess the next step would be to test a few more configs... depending on what VIF you set up. I'm not sure how test 2 increased drastically over test 1 if both of them are transfering over 1 physical NIC. It appears that the ndmpd preferred interface has confused the trasnfer by using the VIF. I could potentially see how that could happen. Maybe the VIF kept flipping between the physical NIC you wasn't sure which physical NIC to transfer over under the VIF.

A few additional ideas may be to try....

1. Can you add multiple NDMP interfaces from the same device? if so therefore you could utilise 2 physical NetApp NICs which are not VIFed. So e0c assigned with 1 IP address and e0d assigned with 1 IP address (although I'm guessing this would need multiple volumes to stream rather than just the 1 volume!?). Eitherway this method would avoid the confusion of the Logical VIF layer and trying to negotiate the path.
2. You could try using the VIF with and without the prefered path BUT using IP alias although I think that may confused the backup exec. Within the VIF, you should use a IP address for the number of physical ports within that VIF (if that makes sense). So similar to the first suggestion but this time without being physical port bound, it is logicall assigned with the VIF layer (again, if that mayes any level of sense).

This is typical when using iSCSI - you would usually leave the NICs as Physical NICs rather than having a VIF. So a dedicated IP addresses per physical NIC...  then map them to the host and then let the host handle the MPIO (whether its using Microsoft iSCSI initiator or something similar). When you use VIF's (Multimode/LACP) you have the NetApp controller or Switch trying to do some element of multipathing at the physical NIC layer AND the host is also trying to do some software multipathing and therefore confusion may occur. In this scenario you might think you don't have failover resiliency but if you had enough ports, you could use 4 physical nics. Create two pairs of single mode vifs, this would give you the resiliency AND MPIO at host layer.
This may be going off topic, but I'm trying to understand why you'd get worse performance from the VIF with preferece ndmp interface on over a phsyical NIC or a VIF without prefered ndmp interface on. Maybe it's getting confused and packets are then dropping and so on.

What are your thoughts? have you got any additional configurations your looking to test?

Adrian

JRAYKOWSKI
12,886 Views

Adrian,

I guess we should lay it out a bit better to reduce confusion

Backup Exec 2010R3, FAS-2240 running Data OnTap 8.1.1 and is configured as a NDMP Device in Backup Exec using Ethernet connectivity.

Test 1

configured e0d on PCM assigned an IP and configured e0d as ndmpd.preferred_interface. Transfered 494.9GB averaging 668MB/s - duration 24hrs & 2min to complete.

 

Test2

Configured e0c and e0d as a vif, trunk mode multiple, on PCM assigned an IP. Left e0d as ndmpd.preferred_interface.  Transfered 37.8GB averaging 2180MB/min - duration 15mins

 

Test3

Configured e0c and e0d as a vif, trunk mode multiple, on PCM assigned an IP. Configured vif as ndmpd.preferred_interface.  Transferred ONLY 11,805,696 bytes in 7 hours

   

  I can add multiple NDMP Devices to Backup Exec but I cannot see how I can do it in OnTap.  I could configure the second PCM as a production filer, it is currently the secondary node in a HA Pair, not really what I want to do.  We have a Flash Pool configured and I would assume leave it alone if I could  I have tried the IP alias options before and Backup Exec did not like it at all.  I would like to create the two pair VIFs but do not have enough ports so I am at a loss for more options.

  At this point I do not know what else I can try.  I like the results of option 2 but will it be reliable?  What happens when I upgrade to 8.1.2?  Will performance be the same or change and will I have to go back through this drill again to find the best configuration?

  Your thoughts?  Anything else you might like to try?

 

Jim

adrianlowe
10,376 Views

Hello Jim,

I think you've done a great job with testing so far. Although I wouldn't perform tests on production data.

I agree test 2does looks the best with regards to transfer rates. It is also a VIF so technical it should failover to e0c IF e0d is not available (e0d should just be its preferred route if both ports are up - given it is the favoured interface). The only one way to test this is to actually pull the cable or down the port during transfer or prior to doing the large transfer, try doing a smaller sized backup and monitor the port activity on the netapp (using the ifstat -acommand) to see which port the data is actually transfering through. I would assume with the "test 2" config... you would see all the ndmp backup data transferring via the e0d port within the vif3... then IF and when you wanted to test the port failover, you could pull the e0d port out to see if the rest of the data continues transferring via the e0c port (also in the VIF).

Note: You don't need to transfer data to test this, you could do a continuous pingto the IP address of that VIF (vif3) (although I'm not sure whether you will see much activity generated from a continuous ping within the ifstat -a output) but you should see the pings drop a few whilst the vif re-negotiates and then the pings should continue like it should during a HA failover too -which reminds me to tell you, make sure you assign the partner failover addresses too thats another test you could test but not whilst in production although those now we're testing resiliency over performance)

Regarding 8.1.2 I don't think it will impact anything with regards to NDMP tape backups (touch wood there are no bugs, although 8.1.2 does seem to be the most stable GA release of ONTAP at the moment). For more information you can visit: http://support.netapp.com/NOW/download/software/ontap/8.1.2/ and read the release notes... there any many new additional features such as 400GB support on FAS22xx series systems, ONTAP Edge-T app and 10GB support on FAS2220 and a few more. with regards to NDMP there are quiet a few changes since 8.0.1 which I wasn't aware of.

  • Support for designated range of ports for NDMP data connections (ONTAP 8.0.1)
  • Support for restore to volumes in 64-bit aggregates using SMTape engine (ONTAP 8.0.1)
  • Support for incremental and differential backups using SMTape engine (ONTAP 8.1.1)

Not that you need any of the above, but it's nice to be aware of

    

If you look at the ONTAP 8.1.2 release notes... and do a search for NDMP you will be able to read in greater detail.

Let me know how you get on if you manage to perform any more tests for the above.

I hope you end up with a reasonably fast transfer rate with full resiliency at both port level and controller failover level back to your server.

Adrian

JRAYKOWSKI
10,376 Views

Adrian,

  I am going to run another test backup using config 2 and then pull the cable and see what happens.  Regardless of the ultimate config I will need to configure both PCMs the same, correct?

Jim

adrianlowe
10,376 Views

Indeed, you would ideally want both controllers with similar config in the event of failover. You can configured them to host to different servers initially but as long as the controller02 e0c+e0d vif has a partner failover to controller01 AND controller01 ports have access to the same network then the failover should work.

You can run a tool called "Config Advisor" which was previously known as "WireGauge" (originally used for checking cabling to disk shelves) but now is getting more advance with incorporating another NetApp Utility tools called "HA Configuration Checker". This will provide you output of any issues/differences between configs between both controllers  i.e. it will compare license keys/and the RC file configs to ensure you have the ability to failover prior to actually testing the physical failover.

You should be able to access all these tools from the NetApp Utility ToolChest: http://support.netapp.com/eservice/toolchest

Search:

5.Config Advisor (WireGauge renamed) 
14.HA Configuration Checker (cf-config-check.cgi)

I hope these are of some use to you.

Adrian

JRAYKOWSKI
10,376 Views

Adrian,

I started another backup, pulled e0d, put it back, then pulled e0c, and put it back and in all cases the backup dropped back down to about 600MB/min and then rose again to ~2200MB/min for backups.  So that with that testing said and done I think we have a final production configuration and it look like this:

FAS-2240 in a HA Pair

 

PCM-1 (Primary Node)

aggr0 - 3 SAS Disks and 1 spare supporting rg0

rg0 -  SAS Disks supporting vol0

vol0 - Data OnTap and Boot Volume

 

fp_aggr - 16 SAS Disks and 1 spare plus 5 SSD Disks and 1 spare

rg0 - SAS Disks supporting Volumes View_Dedicated, View_Linked, View_Templates, vol_vm_1, and vol_home

rg1 - SSD Disks supporting the Flash Pool capability

 

e0a and e0b assigned to VIF1 on the production network supporting CIFS Shares partnered to VIF2

e0c and e0d assigned to VIF3 on the backup network support CIFS backups to Backup Exec  partnered to VIF4

 

e0m on the management network partnered to PCM-2 e0m

e0p on the management network

 

FC

1a and 1b - Connected to dual Cisco MDS-9124 FC Switches presenting LUNs View_Dedicated, View_Linked, View_Templates, vol_vm_1

 

CIFS -  using VIF1 and presenting vol_home

PCM-2 (Secondary Node)

aggr0 - 3 SAS Disks and 1 spare supporting rg0

rg0 -  SAS Disks supporting vol0

vol0 - Data OnTap and Boot Volume

e0a and e0b assigned to VIF2 on the production network supporting CIFS Shares  partnered to VIF1

e0c and e0d assigned to VIF4 on the backup network support CIFS backups to Backup Exec  partnered to VIF3

e0m on the management network  partnered to PCM-1 e0m

e0p on the management network

 

FC

1a and 1b - Connected to dual Cisco MDS-9124 FC Switches

 

CIFS -  using VIF2

  Having not touched a NetApp in about 10 years, working off and on for almost 6 weeks, reading paper upon paper, and with much help I learned that it is some times a bit of a pain coming back to a technology you haven't used in a while. 

Thanks for all your help. If you have any questions please let me know,

Jim

Public