Bacula-users

Re: [Bacula-users] Excruciatingly slow backup to Tape - local and remote

2012-12-17 19:53:19
Subject: Re: [Bacula-users] Excruciatingly slow backup to Tape - local and remote
From: Steve Ellis <ellis AT brouhaha DOT com>
To: bacula-users AT lists.sourceforge DOT net
Date: Mon, 17 Dec 2012 16:50:12 -0800
You don't mention the technology behind your tape drive, database backend, CPU, RAM, or what your disk subsystem looks like--all of which would be useful to have a reasonable chance to analyze even vaguely properly, but I'll wade in nonetheless.

You are almost certainly shoeshining the heck out of your tape media.  I believe most drive types will have to resort to stopping and starting once you get below a certain threshold (for LTO3, I think this is somewhere around 30MB/sec, as an example), the tape drive will wind up starting a stopping very frequently (as well as rewinding a bit).  I imagine this will wear out both tapes and drive mechanisms pretty quickly.

Since you are not spooling data, I suspect you also are not spooling attributes--my first guess would be that your database backend is insufficiently fast--at a minimum you should spool attributes, and spooling data on even the local backups can yield performance gains, depending on your disk subsystem architecture.   If network and disk subsystem can't deliver at least the minimum streaming rate for your tape drive, you need to spool to a fast enough disk to keep up with the tape drives needs.

If you are using something like LTO5, I think spooling using either SSD or RAID0 may almost be required--but perhaps LTO5 actually supports relatively low minimum streaming rates--I don't know, as I can't afford LTO5 (and I think the minimum streaming rate may well be manufacturer dependent).

-se


On Mon, Dec 17, 2012 at 3:47 PM, shockwavecs <bacula-forum AT backupcentral DOT com> wrote:
I'm only seeing around 2MB/s write speed to tape. I have read all about spooling the data, etc. I would expect this to be an issue if I was complaining about getting 18-25 MB/s instead of 90-100MB/s maybe? Anyways, I simply cannot understand why a single job runs at 2 MB/s and 3 jobs will run about 2.5MB/s total. See output below (all servers in the same rack/switch running at the same time):

Writing: Full Backup job MSSERVER2 JobId=24 Volume="BAL_01"
    pool="Default" device="AIT-4" (/dev/nst0)
    spooling=0 despooling=0 despool_wait=0
    Files=18,121 Bytes=9,726,633,719 Bytes/sec=1,008,254
    FDReadSeqNo=305,615 in_msg=251883 out_msg=5 fd=6

Writing: Full Backup job TS JobId=25 Volume="BAL_01"
    pool="Default" device="AIT-4" (/dev/nst0)
    spooling=0 despooling=0 despool_wait=0
    Files=20,896 Bytes=5,088,180,026 Bytes/sec=691,234
    FDReadSeqNo=256,355 in_msg=195224 out_msg=5 fd=8

Writing: Full Backup job BALDC1 JobId=26 Volume="BAL_01"
    pool="Default" device="AIT-4" (/dev/nst0)
    spooling=0 despooling=0 despool_wait=0
    Files=39,608 Bytes=4,514,041,685 Bytes/sec=658,503
    FDReadSeqNo=411,027 in_msg=295248 out_msg=5 fd=10



Here is the output of backing up the bacula server itself (this proves spooling would not help me, right?):


  Build OS:               i686-pc-linux-gnu redhat
  JobId:                  12
  Backup Level:           Full (upgraded from Incremental)
  Client:                 "bacula-fd" 5.2.6 (21Feb12) i686-pc-linux-gnu,redhat,
  FileSet:                "Full Set" 2012-12-17 11:24:20
  Pool:                   "Default" (From Job resource)
  Catalog:                "MyCatalog" (From Client resource)
  Storage:                "SONY" (From command line)
  Scheduled time:         17-Dec-2012 11:24:18
  Start time:             17-Dec-2012 11:24:22
  End time:               17-Dec-2012 12:34:46
  Elapsed time:           1 hour 10 mins 24 secs
  Priority:               10
  FD Files Written:       118,347
  SD Files Written:       118,347
  FD Bytes Written:       8,199,458,340 (8.199 GB)
  SD Bytes Written:       8,215,432,405 (8.215 GB)
  Rate:                   1941.2 KB/s
  Software Compression:   None
  VSS:                    no
  Encryption:             no
  Accurate:               no
  Volume name(s):         BAL_01
  Volume Session Id:      1
  Volume Session Time:    1355761332
  Last Volume Bytes:      8,225,667,072 (8.225 GB)
  Non-fatal FD errors:    0
  SD Errors:              0
  FD termination status:  OK
  SD termination status:  OK
  Termination:            Backup OK
 Begin pruning Jobs older than 6 months .
 No Jobs found to prune.
 Begin pruning Files.
 No Files found to prune.
 End auto prune.

+----------------------------------------------------------------------
|This was sent by shockwavecs AT gmail DOT com via Backup Central.
|Forward SPAM to abuse AT backupcentral DOT com.
+----------------------------------------------------------------------



------------------------------------------------------------------------------
LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial
Remotely access PCs and mobile devices and provide instant support
Improve your efficiency, and focus on delivering more value-add services
Discover what IT Professionals Know. Rescue delivers
http://p.sf.net/sfu/logmein_12329d2d
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users

------------------------------------------------------------------------------
LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial
Remotely access PCs and mobile devices and provide instant support
Improve your efficiency, and focus on delivering more value-add services
Discover what IT Professionals Know. Rescue delivers
http://p.sf.net/sfu/logmein_12329d2d
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users
<Prev in Thread] Current Thread [Next in Thread>