Forum Discussion

v-2kisad's avatar
v-2kisad
Level 2
8 years ago

Slow Restore to Unix Client

Restore job has been running for over 30hrs however the ammount of data restored  is 250 GB. The backup job had taken 3-4 hrs and expected restore size is 350GB. I can see wait & delay message in the activity monitor job.Is there anything that we can do increase the speed or the job and also find out why it is taking so long.

Thanks is advance.


Netbackup Master & Media : 7.7.3 on Windows 2008 R2
Restore Client: Cent OS

05/30/2017 13:36:01 - Info bprd (pid=6588) Searched (10,283,852) files of (10,393,901) files for Restore Job ID 4206616.xxx
05/30/2017 13:40:50 - started process bptm (pid=8932)
05/30/2017 13:40:56 - begin reading
05/30/2017 13:42:48 - Info bptm (pid=9528) waited for empty buffer 1642 times, delayed 3249 times
05/30/2017 13:42:51 - begin reading
05/30/2017 13:43:14 - Info bptm (pid=9528) waited for empty buffer 25 times, delayed 1489 times
05/30/2017 13:43:17 - begin reading
05/30/2017 15:51:07 - Info bptm (pid=9528) waited for empty buffer 13722 times, delayed 487687 times
05/30/2017 15:51:07 - end reading; read time: 2:07:50
05/30/2017 15:51:11 - begin reading
05/30/2017 17:03:50 - Info bptm (pid=9528) waited for empty buffer 5397 times, delayed 283417 times
05/30/2017 17:03:53 - begin reading
05/30/2017 19:46:33 - Info bptm (pid=9528) waited for empty buffer 13479 times, delayed 629246 times
05/30/2017 19:46:34 - end reading; read time: 2:42:41
05/30/2017 19:46:37 - begin reading
05/30/2017 23:21:14 - Info bptm (pid=9528) waited for empty buffer 12630 times, delayed 848089 times
05/30/2017 23:21:17 - begin reading
05/31/2017 02:29:19 - Info bptm (pid=9528) waited for empty buffer 18284 times, delayed 726346 times
05/31/2017 02:29:19 - end reading; read time: 3:08:02
05/31/2017 02:29:23 - begin reading
05/31/2017 04:49:26 - Info bptm (pid=9528) waited for empty buffer 9418 times, delayed 547928 times
05/31/2017 04:49:26 - end reading; read time: 2:20:03
05/31/2017 04:49:30 - begin reading
05/31/2017 15:05:07 - Info bptm (pid=9528) waited for empty buffer 37448 times, delayed 2418900 times
05/31/2017 15:05:07 - end reading; read time: 10:15:37
05/31/2017 15:05:11 - begin reading



3 Replies

  • We have often seen that mismatched NIC and/or switch settings (auto vs fixed) will give acceptable backup performance but poor restore rates.

    Try to test network transfer using ftp from media server to client with a file that is at least 1GB.
    • Michael_G_Ander's avatar
      Michael_G_Ander
      Level 6

      I would also look at the receiving file system/disk as it seems that the data cannot be delivered fast enough.

      In general reading is faster than writing, but usually not by a factor 10.

      Some times increasing the network communication buffer on the client can improve restore speed.

      Of course memory & cpu on the client can be limiting factors too.

       

      • ElGringo's avatar
        ElGringo
        Level 6

        Hi,

        WHat is the restore policy type ? If it is VMware policy type do you try to do a file level restore ?

        Furthermore, I see a lot of waited for empty data buffer massage. So, I guess network or client can be a limited factor. But it may NOT be the only one ! Have you tried to run an iperf test between the media and client ?