'Phil Stracchino' wrote:
>On 10/07/10 13:47, Roy Sigurd Karlsbakk wrote:
>> Hi all
>>
>> I'm planning a Bacula setup with ZFS on the SDs (media being disk,
>> not tape), and I just wonder - should I use a smaller recordsize (aka
>> largest block size) than the default setting of 128kB?
>
>Actually, there are arguments in favor of using a larger, not smaller,
>block size for applications such as this where you expect the major
>usage to be extended streaming reads and writes.
As a general rule I do agree (especially when dealing with sequential
I/O) - but it still is dependend on the application I/O.
>That said, my own disk SD runs on ZFS with default block size and works
>just fine.
$ for f in `ls`; do zfs get -Hpo value recordsize storage01-01/bacula/$f; done
| uniq -c
189 131072
189 ZFS filesystems all with the default 128k recordsize.
A quick peek with dtrace shows this :
$ dtrace -n 'sysinfo:::writech / execname == "bacula-sd" / {
@dist[execname] = quantize(arg0); }'
dtrace: description 'sysinfo:::writech ' matched 4 probes
^C
bacula-sd
value ------------- Distribution ------------- count
2 | 0
4 | 4
8 | 0
16 | 3
32 | 0
64 |@@@@ 75216
128 |@ 18477
256 |@@@@ 74357
512 | 0
1024 | 0
2048 | 0
4096 | 0
8192 | 0
16384 | 0
32768 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ 514260
65536 | 0
This was taken during a single full backup of a windows client.
The sysinfo:::writech call covers all write(2), writev(2) or pwrite(2)
system calls - writes generated by the bacula-sd process seem to be
limited to 32k, regardless of the underlying recordsize (upper block
size limit).
I'll run this tonight when we have ~100 clients backing up towards this
machine - I'll will monitor the actual I/O size as seen by ZFS aswell
and post the output is someone is interested ...
>
>--
> Phil Stracchino, CDK#2 DoD#299792458 ICBM: 43.5607, -71.355
> alaric AT caerllewys DOT net alaric AT metrocast DOT net phil AT
> co.ordinate DOT org
> Renaissance Man, Unix ronin, Perl hacker, Free Stater
> It's not the years, it's the mileage.
>
>------------------------------------------------------------------------------
>Beautiful is writing same markup. Internet Explorer 9 supports
>standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 & L3.
>Spend less time writing and rewriting code and more time creating great
>experiences on the web. Be a part of the beta today.
>http://p.sf.net/sfu/beautyoftheweb
>_______________________________________________
>Bacula-users mailing list
>Bacula-users AT lists.sourceforge DOT net
>https://lists.sourceforge.net/lists/listinfo/bacula-users
--
Med venlig hilsen / Best Regards
Henrik Johansen
henrik AT scannet DOT dk
Tlf. 75 53 35 00
ScanNet Group
A/S ScanNet
------------------------------------------------------------------------------
Beautiful is writing same markup. Internet Explorer 9 supports
standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 & L3.
Spend less time writing and rewriting code and more time creating great
experiences on the web. Be a part of the beta today.
http://p.sf.net/sfu/beautyoftheweb
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users
|