fbackup over network is slooooow

[Randy Hagedorn]

Hi,
We have been experimenting with fbackup from one k460 to another K460 over the network. When we fbackup our production system to it's own DDS-3 internal tape drive it takes about (1) hour. When the same fbackup is performed over the network to another K460 on an identicate tape drive, it takes (7) hours. Both systems are using 100M LAN autonegotiate Full Duplex.

We know that fbackup over a network will be slower, but is it reasonable that it would be seven-times slower? Does anyone think that there may be a system tuning problem or should we be looking at a 3rd-party backup product?

Thanks in advance ... Dewy

[harry d brown jr]

Sure it's reasonable, but I agree that it's not acceptable. How much data are you backing up?

[Sachin Patel · Accepted solution]

Hi
We have one nfs server with 50gb data takes around 4 days through network for backup.

Good news. We just found new way to fast this process. Now it take only 24 hour only.

Create file called config.backup
blocksperrecord 256
records 32
checkpointfreq 4096
readerprocesses 6
maxretries 5
retrylimit 5000000
maxvoluses 100
filesperfsm 800

use that with fbackup
#fbackup -c config.backup -g -0v -f server:

Sachin

[A. Clay Stephenson]

One thing to check is that you may not actually be running 100FD. Auto-negotiate is not always reliable and one or both of the boxes may have fallen back to something less. I would hard set to 100FD on both the K-boxes and the switches to eliminate that possibility.

Plan B: Compress the data before sending it over the network, something like this:

fbackup -f - | compress | remsh boxB dd of=xxx

I suspect that you are not getting the full 100FD because OmniBack (which is really a good choice for this) does not suffer degradations that big over the network.

[Randy Hagedorn]

Harry,

We are backing up about 5.9 Gig. which we feel isn't really big.

Thanks, Dewy

[harry d brown jr]

DuWayne,

I agree, ~6gb is nothing. Sachin made some great points. Also, using lanadmin, check your lan card to make sure you aren't getting excessive errors.

[Paula J Frazer-Campbell]

Hi
Try this:-

NFS export the data areas
NFS import them on the backup server

If you have room create a data area on the backup server.

copy the data from export machine to backup machine:-
cd to mounted data area:-
time find . | cpio -pudvm /

I move approx 12 Gig of data on a 100 meg network in about 1.5 hours by this method.

This will allow to to check if it is you network that is slowing you down.

HTH

Paula

[George C. Abraham_2]

have you loaded the latest N/W pathes from HP. We had a similar prblem with a L2000 After loading the patches the performance has improved a bit by around 15%

regards

[Justo Exposito]

Hi,

I had this problem and then I tried to change the configuration parameters of fbackup for my DLT tape over the network and with this values I obtained the best performance:

blocksperrecord 1024
records 16
checkpointfreq 256
readerprocesses 2
maxretries 5
retrylimit 5000000
maxvoluses 100
filesperfsm 200

Perhaps you must try to do a fbackup with different params (records and blocksperrecords) and see the results, in my case this change from 4 hours to 2 hours.

good luck.

Justo.

[Bill Hassell]

This definitely sounds like a faile auto-negotiation, so your LAN speed is really about 2Mbit/sec, not 100mbit/sec. Verify this with lanadmin and if it reports 100Mbit at half-duplex, the negotiation failed and the interface is dropped to half-duplex (as defined by the standard).

Autonegotiation fails when the cable is a specific length (25-50 meters) or with certain combinations of LAN chip vendors. Lock the switch and your LAN card to 100Mbit and look at lanadmin stats again.