ZFS Compression Oddities with Amanda

2013-05-01 Thread Guy Sisalli
I'm attempting to commit 7 TB of text to tape. It's presently stored in
a natively gzip-9 compressed zvol, weighing in at 1.7 TB.

My holding area is 5 TB, and is set to a native gzip-5 compression.

The functional difference between gzip-5 and gzip-9 is not very much:
Level 9 compression has a 4-8% advantage over level 5. The entire DLE
(taken as files, not a snapshot) should fit quite comfortably in my
holding area. It didn't!

I watched my holding area balloon to 4 TB and keep right on going, as if
it wasn't compressing at all. Is there any scenario in which this might
happen? Would you recommend against a setup like the one I've described?
I'm happy to offer any details needed, but this is probably a good start:

Source:

NAMEPROPERTY  VALUE  SOURCE
zulu01/keyRepo  type  filesystem -
zulu01/keyRepo  creation  Tue Jan  8 13:48 2013  -
zulu01/keyRepo  used  1.74T  -
zulu01/keyRepo  available 6.60T  -
zulu01/keyRepo  referenced1.74T  -
zulu01/keyRepo  compressratio 4.65x  -
zulu01/keyRepo  mounted   yes-
zulu01/keyRepo  quota none   default
zulu01/keyRepo  reservation   none   default
zulu01/keyRepo  recordsize128K   default
zulu01/keyRepo  mountpoint/tank/datastoredefault
zulu01/keyRepo  sharenfs  offdefault
zulu01/keyRepo  checksum  on default
zulu01/keyRepo  compression   gzip-9 local

Hold:

NAME  PROPERTY  VALUE  SOURCE
hold  type  filesystem -
hold  creation  Fri Dec  7 10:34 2012  -
hold  used  257M   -
hold  available 5.35T  -
hold  referenced243M   -
hold  compressratio 1.00x  -
hold  mounted   no -
hold  quota none   default
hold  reservation   none   default
hold  recordsize128K   default
hold  mountpoint/hold  default
hold  sharenfs  offdefault
hold  checksum  on default
hold  compression   gzip-5 local

-- 
Guy Sisalli gsisa...@mynetwatchman.com
IT Operations Manager
myNetWatchman.com
Atlanta, GA
+1.631.708.8346


Re: ZFS Compression Oddities with Amanda

2013-05-01 Thread Jean-Louis Martineau

Guy,

Are you also using amanda software compression?

Using compression on the holding disk is probably just a waste of CPU as 
the data is compressed once and decompressed once, but if it the only 
way it can fit in the holding disk.


If you use amanda software compression, then the data is compressed on 
holding disk and on tape.


Jean-Louis

On 05/01/2013 01:10 PM, Guy Sisalli wrote:

I'm attempting to commit 7 TB of text to tape. It's presently stored in
a natively gzip-9 compressed zvol, weighing in at 1.7 TB.

My holding area is 5 TB, and is set to a native gzip-5 compression.

The functional difference between gzip-5 and gzip-9 is not very much:
Level 9 compression has a 4-8% advantage over level 5. The entire DLE
(taken as files, not a snapshot) should fit quite comfortably in my
holding area. It didn't!

I watched my holding area balloon to 4 TB and keep right on going, as if
it wasn't compressing at all. Is there any scenario in which this might
happen? Would you recommend against a setup like the one I've described?
I'm happy to offer any details needed, but this is probably a good start:

Source:

NAMEPROPERTY  VALUE  SOURCE
zulu01/keyRepo  type  filesystem -
zulu01/keyRepo  creation  Tue Jan  8 13:48 2013  -
zulu01/keyRepo  used  1.74T  -
zulu01/keyRepo  available 6.60T  -
zulu01/keyRepo  referenced1.74T  -
zulu01/keyRepo  compressratio 4.65x  -
zulu01/keyRepo  mounted   yes-
zulu01/keyRepo  quota none   default
zulu01/keyRepo  reservation   none   default
zulu01/keyRepo  recordsize128K   default
zulu01/keyRepo  mountpoint/tank/datastoredefault
zulu01/keyRepo  sharenfs  offdefault
zulu01/keyRepo  checksum  on default
zulu01/keyRepo  compression   gzip-9 local

Hold:

NAME  PROPERTY  VALUE  SOURCE
hold  type  filesystem -
hold  creation  Fri Dec  7 10:34 2012  -
hold  used  257M   -
hold  available 5.35T  -
hold  referenced243M   -
hold  compressratio 1.00x  -
hold  mounted   no -
hold  quota none   default
hold  reservation   none   default
hold  recordsize128K   default
hold  mountpoint/hold  default
hold  sharenfs  offdefault
hold  checksum  on default
hold  compression   gzip-5 local





Re: ZFS Compression Oddities with Amanda

2013-05-01 Thread Brian Cuttler

I'd thought for both ZFS and the newer (LTO) tape drives
that HW-compression was determined on a block by block
basic (if enabled) so that expansion of data would not occur.

Granted, this does nothing to help with on CPU usage, but I'd
thought it did save, rather, preserve, storage volume.

On Wed, May 01, 2013 at 01:41:35PM -0400, Jean-Louis Martineau wrote:
 Guy,
 
 Are you also using amanda software compression?
 
 Using compression on the holding disk is probably just a waste of CPU as 
 the data is compressed once and decompressed once, but if it the only 
 way it can fit in the holding disk.
 
 If you use amanda software compression, then the data is compressed on 
 holding disk and on tape.
 
 Jean-Louis
 
 On 05/01/2013 01:10 PM, Guy Sisalli wrote:
 I'm attempting to commit 7 TB of text to tape. It's presently stored in
 a natively gzip-9 compressed zvol, weighing in at 1.7 TB.
 
 My holding area is 5 TB, and is set to a native gzip-5 compression.
 
 The functional difference between gzip-5 and gzip-9 is not very much:
 Level 9 compression has a 4-8% advantage over level 5. The entire DLE
 (taken as files, not a snapshot) should fit quite comfortably in my
 holding area. It didn't!
 
 I watched my holding area balloon to 4 TB and keep right on going, as if
 it wasn't compressing at all. Is there any scenario in which this might
 happen? Would you recommend against a setup like the one I've described?
 I'm happy to offer any details needed, but this is probably a good start:
 
 Source:
 
 NAMEPROPERTY  VALUE  SOURCE
 zulu01/keyRepo  type  filesystem -
 zulu01/keyRepo  creation  Tue Jan  8 13:48 2013  -
 zulu01/keyRepo  used  1.74T  -
 zulu01/keyRepo  available 6.60T  -
 zulu01/keyRepo  referenced1.74T  -
 zulu01/keyRepo  compressratio 4.65x  -
 zulu01/keyRepo  mounted   yes-
 zulu01/keyRepo  quota none   default
 zulu01/keyRepo  reservation   none   default
 zulu01/keyRepo  recordsize128K   default
 zulu01/keyRepo  mountpoint/tank/datastoredefault
 zulu01/keyRepo  sharenfs  offdefault
 zulu01/keyRepo  checksum  on default
 zulu01/keyRepo  compression   gzip-9 local
 
 Hold:
 
 NAME  PROPERTY  VALUE  SOURCE
 hold  type  filesystem -
 hold  creation  Fri Dec  7 10:34 2012  -
 hold  used  257M   -
 hold  available 5.35T  -
 hold  referenced243M   -
 hold  compressratio 1.00x  -
 hold  mounted   no -
 hold  quota none   default
 hold  reservation   none   default
 hold  recordsize128K   default
 hold  mountpoint/hold  default
 hold  sharenfs  offdefault
 hold  checksum  on default
 hold  compression   gzip-5 local
 
 
---
   Brian R Cuttler brian.cutt...@wadsworth.org
   Computer Systems Support(v) 518 486-1697
   Wadsworth Center(f) 518 473-6384
   NYS Department of HealthHelp Desk 518 473-0773