BottleNeck Target 99%

[GarethWilson]

Hi

We have just upgraded to B&R 6.1 from 5.02, when the jobs now run they are about 5x as slow as they were under v5, in the stats of the job its showing Bottleneck: Taget, which i take it is the repository that it is writing to ?

Our target is a HP lefthand iscsi san and we had no issues with any speed writing to this before upgrading to v6.1 if i look at the stats of the ethernet card going to the iscsi backup device it is anything from 1% to 15% network utilisation ?

What could be wrong as backup jobs that would run before in 30 mins are now taking like 4 hours to complete and are going at about 11mb/s where they used to go in the 100's

Help !

[foggy]

Bottleneck target shows that the job is waiting for the target storage to write the data most of the time, which results in overall low processing performance.

Could you please give a bit more details on your setup? Have you performed an in-place upgrade or created additional proxy/repository servers? What backup mode do you use (forward/reverse incremental)?

Also, please review this topic for possible reasons of performance decrease.

[GarethWilson]

Yes we did an inplace upgrade we havent created any additional proxy/repository servers, we are using reverse incremental jobs, the vmware luns are hosted via 4gb fibre channel to 2 x 3PAr F400 disk arrays and the backup is on a HP Lefthand ISCSI San, i will look at the other post

Thanks

Gareth

[hnicaise]

Hi,
We have the same type of problem with Veeam V6.1 since we start using this release. Before (in V5) we never had problems with replications or backup, the now when we try replicate Vm, the first try sound ok, but every incremental take 3 to 7 hours to end ... and always à bottleneck on target 99%.
I've try different way of solution but none seems to solve the problem.

For information I don't upgrade from veeam V5 to Veeam V6.1, it's a full new VM with the same number of vCPU and memory.

[tsightler]

You might want to try forcing the target proxy to NBD mode rather than hotadd. Also, be sure you have proxies on both the source and target side and that they are being properly selected in the job. Manually pick them if you have to.