Backup speeds/Pre-processing 2

[tEkHEd]

Hi there.

I am currently trying to find out why our backups are running slowly.

We use a gigabit backbone to backup to HP818 or HP8000 surestore autoloaders.

Over the last week or so our throughput rates have dropped right down to about 50Mb/m from 300Mb/m.

This is NOT affecting all servers though, which is a surprise as all config's of Backup Exec 8.6 are exactly the same.

I also have an issue with one of our media servers taking an exceptionally long time to complete the Pre-processing section of the jobs. This can take in excess of 20 mins to complete!!

I am unable to identify why this is the case (only on 1 server) when if I run a backup job from a different media server (similar hardware, same exec config) the pre-processing takes a few miliseconds. This test was carried out on the same network volume, at the same time of day.

Can anyone give me any pointers to look at? I thought initially it was to do with the "Use Imaging Option if available" setting in the Exec main options and the backup job options, but this setting was not checked!

I also thought it may have to do with OFO, but this doesn't make sense as I only have a slow pre-processsing with 1 server (compaq ML360, 2 x 733 PIII, 1024Mb RAM, RAID 5, Adaptec 39160 SCSI card/Adaptec 2940 SCSI card, Win2K Server SP 3)

Any help is appreciated

thanks in adv.

 

[justin2000]

tekhed

The pre-processing on the server - OFO looks for quiet time and after a certain amount of time gives up and then just does a normal backup. the defaults are 5 seconds to wait for quiet time and to try for 2000 seconds.
To test, either lower the amount of time it waits to get quiet time, say to 300 seconds, or try a backup without OFO turned on.

And for the slow backup of servers, check for any SCSI timeout errors in the system event log, 7,9,11,15 errors, these will slow your backup down.

 

[tEkHEd]

Hi Justin..

Thanks for that info.. I have had a look in the eventvwr and I am indeed getting an error 9 (timeout) issue.

The device, \Device\SCSI\adpu160m2, did not respond within the timeout period

Do you know how I can resolve this? Can I increase the timeout threshold?

also would you happen to know if HyperThreading has an impact on the throughput?

thanks

 

[tEkHEd]

A fix for SCSI timeouts is available from Adaptec for Netware servers..

This consists of:

1: Disabling Domain Validation
2: Setting Maximum Sync Transfer Rate to that of the DLT device.

I have tried to implement this same change with the Win2K server, but unfortunately this has not had any impact on the error message.

 

[justin2000]

tekhed

A couple of things you can try..

Need to go into this from boot up time Ctrl-A

1. Disable wide negotiation on teh SCSI card for the ID of the tape drive.
2. Try lowering the transfer rate of the ID as well.

 

[tEkHEd]

Justin,

Thanks but that didn't work.. it was kinda the same as the one I mentioned for the Netware ^^^^^^

I think that this may be an issue with the Dell hardware combo.. they have issued a new BIOS (A15) and new adaptec drivers (v6.xxx) these weren't even available on the adaptec site, so I am thinking that this is a compatibility issue because it seems as though this driver from adaptec has been specifically written for poweredge servers (I am not sure about that, but dell have had this version available for a while, and like I said it is not on the adaptec site!!)

gonna try that stuff on monday.. but have a star anyway, cos ur post was a good solution