Fluctuating backup performance

[golilang]

I have been having some problems getting 2 new clients to complete their backups successfully on the same day. One or the other will complete. The other is not failing but facing bandwidth issues usually down to 1mb with a total of 4 sessions. With 600+gb each this could go on forever.

I do have all servers on forced full duplex and on gigabit ethernet.

One is set for 22:00 the other 03:00 and no other jobs scheduled at this time. There should be plenty of bandwidth to go around.

Anyone have any ideas?

 

[rgzimmer]

It is probably not be a bandwidth issue with the network if you are running a gigabit backone, but your backup server may not be able to handle the large load.

Networker is very processor intensive. If you are doing a large backup and running directives, the backup server may not be able to handle it. Your processor is likely at 100% the entire time. Memory is usually not an issue and on my server remains constant throughout all backups.

Make sure that your backup server has at least two or more processors running as fast a bus speed as possible, and that each of your tape drives is on its own scsi channel.

 

[golilang]

We built the backup server with this in mind so we have a E4500 running solaris 7 with 8x400mhz processors 8gb of ram. We are using 12 tape drives on a StorageTek L700 2 drives per i/o board.

This server is acting as the storage node and backup server.

Networker is on 6.1.1 build 238 as well as the 2 clients that I am having problems with.

 

[rgzimmer]

Looks like plenty of power and then some on there. Have you noticed that the same drives are in use one the slower backup? I had a random problem that turned out to be a failing drive. It just took a couple of days to realize that the slowed backups were always writing to the same drive. It wasn't always the same one each day.

 

[Guest_imported]

Sorry, but my English is not the best.

Is your Legato Networker Server running on a RAID 5 or 6 Array? If yes, it could be the Problem of your bandwidth issues.
Never use a Raid 5 or RAID 6 as a save destination for Networker Files like the File Index. Backupjobs and Indexchecks needs longer.

Dont Ask me why, its hard for me to explain it

In the Legato Networker Version 6.1.1 File Index Checks starts automaticly.

Good Luck!
I hope i could help you

 

[cptb]

Hmm, RAID 5 is slow on writings because of Controlbit calculations. However I hardly think thats your problem. You didn't say anything about your client O/S but If Windows you might have problems with a fragmented FileSystem. Also make sure you don't have any autonegotiation set to On. On CPQ and Intel Nics thats an option under advanced option. Check that again. Also try to ftp the data instead of running Backup. If that is slow too, it's a Network issue if it's much faster it's a Networker or filesystem issue.

 

[Rexxy]

Quick note, Networker is fully dependant on it's thru put with the nsrd process, basically putting it down as a single process stream software, therefore it will only ever utilise ONE processor at any given time. So the other 7 processors wouldnt be used for nsr. Nice ammount of Memory though.....

 

[Guest_imported]

We have the same configuration (E4500 & L700) and have found that it points to the slots the SCSI controllers are in.

Our UNIX support team talked to SUN and were given a diagram which confirmed that the slots are different speeds, I/O bus & 32/64/128bit. The 3 middle slots are the fastest channels.

We can achieve close to 20Mb/s on the fast channels but only 1 - 2Mb/s on others

We are in the process of restructuring the Silo so that we will have 10 drives each with there own SCSI channel.

 

[golilang]

Thanks everyone. I resolved this sometime ago. I noticed that backup speeds were bad only on 2 drives. At first I assumed it might be the HBA since the drives were on the same bus. We have 2 drives per bus. I had StorageTek come out and take a look at the jukebox and we found some bad scsi cables. Replaced them and back in business.

Thanks.