05-09-2013 01:57 AM
Good morning all,
i'm facing a strange behaviour on two servers, w2k8 that are pretty the same, doing same thing, same hw and whatever, that are backing up same SAP db (via brbackup), that are pretty same size, more or less.
Important premises is that they normally from a while ran without any differences, or problems on the troughput.
From some time, and i can't imagine anything changed on their configuration or other networ/devices or whatever around, i see big differences.
Here the log for both.
FAST server (filtered only what does make sense):
5/8/2013 10:00:42 AM - Info bpbrm(pid=18894) INF - Client read timeout = 9600
5/8/2013 10:00:42 AM - Info bpbrm(pid=18894) start bpbkar on client
5/8/2013 10:00:46 AM - Info bptm(pid=19205) using 262144 data buffer size
5/8/2013 10:00:49 AM - Info bptm(pid=19205) setting receive network buffer to 262144 bytes
5/8/2013 10:00:49 AM - Info bptm(pid=19205) using 128 data buffers
5/8/2013 10:00:50 AM - Info bptm(pid=19205) start backup
5/8/2013 10:01:01 AM - begin writing
5/8/2013 11:02:40 AM - Info bphdb(pid=6364) done. status: 0
5/8/2013 11:02:42 AM - Info bptm(pid=19205) waited for full buffer 161943 times, delayed 224203 times
5/8/2013 11:02:44 AM - Info master(pid=19205) StorageServer=PureDisk:master; Report=PDDO Stats for (master): scanned: 88129768 KB, CR sent: 2022655 KB, CR sent over FC: 0 KB, dedup: 97.7%
SLOW one:
5/8/2013 10:00:48 AM - Info bpbrm(pid=19081) INF - Client read timeout = 9600
5/8/2013 10:00:49 AM - Info bpbrm(pid=19081) start bpbkar on client
5/8/2013 10:00:52 AM - Info bptm(pid=19341) using 262144 data buffer size
5/8/2013 10:00:55 AM - Info bptm(pid=19341) setting receive network buffer to 262144 bytes
5/8/2013 10:00:55 AM - Info bptm(pid=19341) using 128 data buffers
5/8/2013 10:00:55 AM - Info bptm(pid=19341) start backup
5/8/2013 10:01:05 AM - Info bptm(pid=19341) backup child process is pid 19740
5/8/2013 11:20:57 AM - Info bptm(pid=19341) waited for full buffer 124192 times, delayed 299685 times
5/8/2013 11:20:58 AM - Info master(pid=19341) StorageServer=PureDisk:master; Report=PDDO Stats for (master): scanned: 71588105 KB, CR sent: 1311886 KB, CR sent over FC: 0 KB, dedup: 98.2%
I can also say that i've the suggested value for w2k8 registry:
TcpWindowSize:DWORD:65535
Tcp4 Checksum Offload disabled
And again, despite one is something smaller than other, the small one is going very low on troughput, near an half, and they are backde up at the same time with same load on the agent and load balance on media:
Backup Active 11256 slower one
Backup Active 10603 slower one
Backup Active 10847 slower one
Backup Active 11646 slower one
Backup Active 33827 faster one
Backup Active 20825 faster one
Backup Active 22013 faster one
Backup Active 19496 faster one
Sum of troughtput is very far from each other.
It was never so, and i changed anything on the Front-end neither on Back-end.
Any advise will be great.
Thank you.
Kind Regards,
05-13-2013 10:58 AM
Hmm.. seems this is the same thread : https://www-secure.symantec.com/connect/forums/nbu-7505-agent-w2k8-different-troughput-very-similar-servers