Video Screencast Help

Slow Exchange DAG SAN Media server backup

Created: 05 Jul 2014 | 3 comments
captain jack sparrow's picture

NBU 7506 (Windows)

Exchange 2010 4 node DAG with NetApp as iSCSI storage

Tape drives are zoned to exchange servers

Seperate policies are created for respective server with DB mounted

local disk backups are 100mb/s+ ... whereas Exchange SAN backups don't exceed more than 10mb/s... its strange!!!

SAN topology is of 4 gbps

buffer tuning helped local backups to offer max speed... but don't know what could i tune here... Can Rep Director help here?

Also wanted to know how Snapshot policy can help reduce window?

Operating Systems:

Comments 3 CommentsJump to latest comment

sdo's picture

Hi Jack,

1) Does the NetBackup backup job start moving data almost immediately, or is there a long time between going active (i.e. client connected) before data starts moving?

2) Are the tape drives zoned to all four MS Exchange servers?

3) Are the same tape drives zoned to any other servers?

4) Which node has robot control?  One of the MS Exchange nodes, or a NetBackup master / media ?

5) Are there other tape drives in the library?

6) Do the tape drives sit behind a FC bridge or hub (e.g. Quantum libraries have the option of installing a piece of kit that acts like a FC hub (e.g. four FC tape drives can be hung off two SAN facing FC ports)) inside the library?

7) Are NetBackup SSO and library option licenses installed on all four DAG servers?

8) Has NetBackup media server been installed, or just plain client?

9) When you say buffer tuning helped speed up backups to local disk; where was the buffer tuning performed, O/S, MS Exchange or NBU - and - what backup tool performed the faster backups to local disk, and what form did the local disk take (NBU DSU?) or something else?

10) Are the firmware and drivers up to date for the HBAs on the MS Exchange server that face the SAN switches?

11) Are the library and tape drive firmware up to date?

12) Have you logged on to the SAN switches and zeroed the counters and then checked during backups to see if any errors are accumulating on the SAN switch ports that face the MS Exchange servers, and that face the tape library and/or tape drives ?

13) Are the servers and tape library/drives on the same SAN switch?

14) If not, how many SAN switch ISL hops are there between the servers and tape library/drives - AND - how fat/wide are the ISL links/trunks (i.e. hops) between all of the SAN switches involved?

Dave.

captain jack sparrow's picture

1) Does the NetBackup backup job start moving data almost immediately, or is there a long time between going active (i.e. client connected) before data starts moving?

it takes time of loading and mounting but this is hardly 2-5 mins usual tape drive takes for allocation etc. concern is throughput of writing data. data from local disk were fast whereas from SAN volume they were low. NBU buffers doesn't differentiate between souce of data from same client i believe.

2) Are the tape drives zoned to all four MS Exchange servers?

yes

3) Are the same tape drives zoned to any other servers?

no they are only zoned to DAG members who are SAN media servers

4) Which node has robot control?  One of the MS Exchange nodes, or a NetBackup master / media ?

one of exchange node

5) Are there other tape drives in the library?

pls elaborate more on this.

6) Do the tape drives sit behind a FC bridge or hub (e.g. Quantum libraries have the option of installing a piece of kit that acts like a FC hub (e.g. four FC tape drives can be hung off two SAN facing FC ports)) inside the library?

No

7) Are NetBackup SSO and library option licenses installed on all four DAG servers?

yes

8) Has NetBackup media server been installed, or just plain client?

media server. Please read again.. client cannot backup itself.. i mentioned data from local disk was backed up in short span it is only SAN data taking time.

9) When you say buffer tuning helped speed up backups to local disk; where was the buffer tuning performed, O/S, MS Exchange or NBU - and - what backup tool performed the faster backups to local disk, and what form did the local disk take (NBU DSU?) or something else?

i didn't performed backups on disk.. backup happens to tape drive only. SAN media server doesn't write data on disk over SAN.. for this i believe NBU FT is to be used.

10) Are the firmware and drivers up to date for the HBAs on the MS Exchange server that face the SAN switches?

yes

11) Are the library and tape drive firmware up to date?

yes

12) Have you logged on to the SAN switches and zeroed the counters and then checked during backups to see if any errors are accumulating on the SAN switch ports that face the MS Exchange servers, and that face the tape library and/or tape drives ?

traffic is seen on port but is slow.. seeing at counters would only ensure traffic is flown through SAN instead of presuming LAN being utilized. 

13) Are the servers and tape library/drives on the same SAN switch?

yes

 

14) If not, how many SAN switch ISL hops are there between the servers and tape library/drives - AND - how fat/wide are the ISL links/trunks (i.e. hops) between all of the SAN switches involved?

 Cheers !!!

CJS

 

sdo's picture

ok

5) How many tape drives are there in total in the library?

12) Definitely not seeing any increasing errors, or loss links indicators (LLI), on the SAN switch ports that face the tape drive(s), and that face the DAG servers, and that face the storage?

14) So there are no other SAN switches in the fabric?

15) One or two fabrics?

16) Definitely checked the speed of the SAN switch ports that face tape drives, DAG servers and storage array?

17) Are SFP light levels (sfpshow) good on the SAN switch ports that face the tape drives and the DAG servers?  Look at both TX and RX on all of the SAN switch ports that face tape, DAG and storage.

18) Are the tape drive targets (WWPNs), and the SAN disk array targets (WWPNs) zoned to the same initiators (HBA ports, WWPNs) on the DAG servers?

19) Has single initiator zoning been implemented, i.e. only one initiator per SAN zone?