Backup of local netware 6.5 server runs very slow

[MvRijt]

We're running a NetWare 6.5 SP6 (OES) server with a lott of services  (bm38sp5,gw701,gwia,webaccess,gwava4,files,zen7,edir8739,netshield463).
BackupExec 9.1.1158 is being used in combination with the OFO option (for groupwise).
For hardware we use a HP Proliant ML350 G4 with 3Gb mem and 3x72Gb SCSI RAID5 (HP SmartArray 641) drive configuration. The HP Ultrium448 LTO backup unit is attached to a separate Adaptec 2932A Ultra320 SCSI card.

The backup at night runs very slow. Meaning, the backup of the local server runs slow. SYS volume - ave. 150 MB/min, VOL1 volume - ave. 100 MB/min.
A Win2k3 server is also backed up using the BE remote agent. Database dump files are located on this server. The c: drive is backed up with an average speed of 380 MB/min and the d: drive 2100 Mb/min. A pretty big difference compared to the local server ...

I've run a tsatest on the local server and the speed measured is as follows. SYS volume  - between 500 and 800 Mb/min. Turning of netshield and backupexec increases the speed about 200 Mb. The volume VOL1 gives lower results, however still much higher than the averages from the backup jobs.
Netshield, btw, is configured for inbound scanning only.

It seems that the NSS config is ok, looking at the tsatest results.

I want to upgrade BE to 9.2 and use the TSAFS for backin up groupwise instead of the OFO.
I tested this scenario in a vmware test envionment, however still using the OFO. This still doesn't give good results.

Are these delays recognizable to anyone?
Who has tips for faster backups??

Regards,
Mario van Rijt
systems administrator
Muziekcentrum Eindhoven

[criley]

What speeds do you see with Backup Exec if you disable OFO and NetShield?

Have you tried backup-to-disk?

[MvRijt]

You mean disabling OFO in the policy or disabling OFO when BE starts?
Is it possible to disable the OFO when issueing the bestart.ncf command, even if it is installed?

No, we haven't tried backup-to-disk.

Regards,
Mario van Rijt

[criley]

---

@MvRijt wrote:
You mean disabling OFO in the policy or disabling OFO when BE starts?

Doesn't really matter which one to be honest.

Is it possible to disable the OFO when issueing the bestart.ncf command, even if it is installed?

Yes, just remove the -to switch from BESTART.NCF which is in SYS:\SYSTEM.

No, we haven't tried backup-to-disk.

May be worth a try to see what speeds you get.

---

[MvRijt]

Nice to know how to disable OFO.
However, only when the OFO is enabled I can see the GroupWise system.
Won't disabling the OFO during backup get in the way of properly backing up the GroupWise environment?

If possible I'll try to make a backup-to-disk so time next week.

[criley]

Not really, you should be using Novell's solutions to properly backup GroupWise - GWTSA.NLM, TSAFSGW.NLM or TSAFS.NLM with the /EnableGW=yes switch. These all ensure consistent backups of GroupWise.

[MvRijt]

I'll keep that in mind. Someone else told me the best options to use are:
- gwava reload
- gwdown,snapshot,gwup (should take only a few minutes according)
You're opinion?

Back to the speed issue.
We had a better backup result last night. We disabled the OFO and also netshield.
On the SYS volume the speed was 470 Mb/min with 45 files in use and 8 skipped.
On the VOL1 volume 177 Mb/min (I hoped more, but could have been worse) with 26 files in use and 17 skipped. This one took about 2 hours and 45 min. to complete. Before this was somewhere between 5 and 6 hours. So ... improvement.

However, every remote server (all windows) show in the job result as not being backed up.
Example:
##NML##Backup set started: 03/28/08 at 04:25:34a
##NML##
##NML##Media ID: 387c0813  Media 1
##NML##Media description: "Maandag - donderdag op 03/28/08 01"
##NML##Set 6 created 03/28/08
##NML##BarCode: LTO_000024
##NML##Set name: "Daily"
##NML##Set description: "Backup van netwerk van maandag t/m donderdag"
##NML##Backup of MFP01/D:
##NML##Drive: "HA:1 ID:3 LUN:0 HP ULTRIUM448 DRV"
##NML##
##NML##  Total directories: 2801
##NML##        Total files: 34362
##NML##        Total bytes: 0
##NML##         Total time: 00:08:57
##NML##         Throughput: 0 bytes/second
##NML##
##NML##Backup set ended: 03/28/08 at 04:35:01a


I looked at the restore area and it shows files (???).
I tried to do a restore of the data and files are actually restored.
So, files were backed up, but it doesn't show in the job result.

Besides disabling the OFO in the policy I also disabled netshield using the policy. Like this:
Pre/Post Options     
Pre/Post Commands                             
Execute before backup: unload netshld
Delay after command (seconds): 30        
Execute after backup: netshld       
Delay after command (seconds): 30        
( )  Execute for every backup set             
(X )  Execute commands once for each data server
Pre/Post Options                              
[ ]  Pre-scan backup sets to estimate size    
[X]  Eject media after job completes          
[ ]  Verify after each backup                 
[X]  Quick Check                              

Like this, the commands are executed for each data server. Meaning, also for the windows servers, right (?)
Could this have anything to do with the inaccurate job results?

[criley]

Well, the Novell solutions allow you to backup without taking down GroupWise so that would always be my recommendation.

As for the speed - it sounds as though either OFO or NetShield (or a combination of the two) are causing a slowdown (my guess is that NetShield is the culprit).

I'm not sure about the other issue, it could be the pre/post command causing this. One way to find out - create 2 NEW jobs that backup a windows box. Have one using these command and one that doesn't - compare the job logs...

[MvRijt]

About the speed. I'm going to perform a backup during the weekend with the OFO disabled and NOT disabling netshield. Let's see if there are differences.

As for the issue remote windows servers showing no throughput in the job log.
I made to separate backup jobs, one using the command and one not.
Using the command I see data being backed up to tape and eventually the job says 0 bytes with 0 throughput.
NOT using the command the job log shows normal.

So, the command is causing this.
Is there any solution for this matter?

[criley]

I've not heard of this problem before..

Basically, the commands are invalid on Windows. Having said that, it obviously should not be causing the job log to report zero bytes/throughput.

A workaround would be to either separate your backups so that you backup your NetWare servers together in one job and your Windows servers in another job or use CRON.NLM to run these commands.

[MvRijt]

If I would separate the jobs, I've got another problem.
Our backup schedule starts at 01.00 AM and has to end (normally) 07.00 AM at the latest.
I don't know exactly how much time it takes to make a full backup of our netware server. And if I would know this, it changes over time.
Is it possible somehow to schedule both backups in such a manner that the backup of the windows servers starts directly after the backup of the netware server has finished??

Last weekend I did a backup without disabling netshield. The SYS volume performed somewhat slower while the VOL1 volume performed somewhat faster. So, hard to tell whether netshield does anything to the backup. Also because netshield is configured to only inbound scanning. Files read from the server are not scanned and the backup only reads files, right (?)

[criley]

It sounds to me that you still need to determine if NetShield is the cause of the slowdown before you go any further with this...

[MvRijt]

This week I've been running every backup job with the OFO *and* netshield disabled.
Next week I'm going to run every backup with only the OFO disabled.

The weekend jobs are different (friday full backup and saterday and sunday differential backups). We've alo experienced problems with the differential jobs. This had something to do with the caching mode of tsafs, so we're testing that issue this weekend.

I haven't done a test using backup-to-disk yet.
I'll probably try that one during business hours with some data not used as much.

[MvRijt]

So, one week later. Here the results.
Netshield disabled
day     SYS     VOL1     POA
mon    490      170        259
thu      472      171        264
wed    517       167       271
thurs   331       161       270
Netshield enabled
day     SYS     VOL1     POA
mon    319      128        265
thu      320      126        261
wed    324       124       265
thurs   333       119       258

So, netshield is making a difference. Looking at the SYS and VOL1 volumes. The POA is going normally, but we're looking at other options for that environment too.
In that case, I'll ask my previous question again:
Is it possible somehow to schedule both backups in such a manner that the backup of the windows servers starts directly after the backup of the netware server has finished??

Still have to perform a backup-to-disk job, btw.

[SpectrumFan]

Novell has released Support Pack 8 for NetWare 6.5 and it fixes the performance problems with external USB drives. More info here: http://www.backupawareness.com/node/6