On Wed, May 4, 2011 at 2:39 PM, Jesper Krogh wrote:
> On 2011-04-28 17:16, Alex Chekholko wrote:
>> Try changing your Maximum Network Buffer size in your bacula-sd config.
>>
>> Something like
>> Maximum Network Buffer Size = 262144 #65536
>> Maximum block size = 262144
>>
>> Keep in mind th
On 2011-04-28 17:16, Alex Chekholko wrote:
> Try changing your Maximum Network Buffer size in your bacula-sd config.
>
> Something like
>Maximum Network Buffer Size = 262144 #65536
>Maximum block size = 262144
>
> Keep in mind that this will make your sd unable to read previous
> backups, I
Martin Simmons:
> > On Fri, 29 Apr 2011 14:29:33 +0200, Dietz Pröpper said:
> > To see wether the file system is indeed the bottleneck you could try
> > to tar the fs to /dev/null and compare the transfer rate to that of
> > your bacula backup.
>
> Good advice, but beware that GNU tar doesn't
> On Fri, 29 Apr 2011 14:29:33 +0200, Dietz Pröpper said:
>
> To see wether the file system is indeed the bottleneck you could try to tar
> the fs to /dev/null and compare the transfer rate to that of your bacula
> backup.
Good advice, but beware that GNU tar doesn't read any files when the
> From: Jason Voorhees
>
>> >
>> > to get the maximum speed with your LTO-5 drive you should enable data
>> > spooling and change the "Maximum File Size" parameter. The spool disk
>> > must be a fast one, especially if you want to run concurrent jobs.
>> > Forget hdparm as benchmark, use bonnie++
Jason Voorhees:
> Well, these are my results of a bonnie++ test:
[...]
> Version 1.03e --Sequential Output-- --Sequential Input-
> --Random- -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
> MachineSize K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP
> /sec %CP
Jason Voorhees:
> On Thu, Apr 28, 2011 at 12:01 PM, John Drescher
wrote:
> >> So do you believe these speeds of my backups are normal? I though my
> >> Library tape with LTO-5 tapes could write at 140 MB/s approx. It
> >> isn't possible to achieve higher speeds?
> >
> > You need to speed up your
Jason Voorhees schrieb:
> > I got the biggest gain by changing "Maximum File Size" to 5 GB. How
> > fast is the disk where you spool file is locatet?
> >
> > A different test would be to create a 10 GB file with data from
> > /dev/urandom in the spool directory and the write this file to tape
> > (
> I got the biggest gain by changing "Maximum File Size" to 5 GB. How
> fast is the disk where you spool file is locatet?
>
> A different test would be to create a 10 GB file with data from
> /dev/urandom in the spool directory and the write this file to tape
> (eg. nst0). Note: this will overwrite
> Ok, I don't have that setting enabled but I could try it. Question:
> how do you decide 5 GB is an optimal value for your LTO-4 tapes? what
> value could I put for my LTO-5 tapes? I don't really understand what
> should be the appropiate value for this directive.
> I don't know how to tell you ho
>
> I got the biggest gain by changing "Maximum File Size" to 5 GB. How
> fast is the disk where you spool file is locatet?
>
Ok, I don't have that setting enabled but I could try it. Question:
how do you decide 5 GB is an optimal value for your LTO-4 tapes? what
value could I put for my LTO-5 tap
Jason Voorhees schrieb:
>
> I think I was confusing some terms. The speed I reported was the total
> elapsed time that my backup took. But now according to your comments I
> got this from my logs:
>
> With spooling enabled:
>
> - Job write elapsed time: 102 MB/s average
> - Despooling elapsed ti
>
> to get the maximum speed with your LTO-5 drive you should enable data
> spooling and change the "Maximum File Size" parameter. The spool disk
> must be a fast one, especially if you want to run concurrent jobs.
> Forget hdparm as benchmark, use bonnie++, tiobench, iozone.
>
> Then after after y
Jason Voorhees schrieb:
>
> I'm running Bacula 5.0.3 in RHEL 6.0 x86_64 with a Library tape IBM
> TS3100 with hardware compression enabled and software (Bacula)
> compression disabled, using LTO-5 tapes. I have a Gigabit Ethernet
> network and iperf tests report me a bandwidth of 112 MB/s.
>
> I'
> I tried to copy a 10 GB file between both servers (Bacula and
> Fileserver) with scp and I got a 48 MB/s speed transfer. Is this why
> my backups are always near to that speed?
>
Try backing up that 10GB file on both servers with bacula.
--
John M. Drescher
--
On 04/28/2011 02:06 PM, Jason Voorhees wrote:
> I tried to copy a 10 GB file between both servers (Bacula and
> Fileserver) with scp and I got a 48 MB/s speed transfer. Is this why
> my backups are always near to that speed?
Try it with "scp -c arcfour" - like compression, encryption introduces
eno
On Thu, Apr 28, 2011 at 3:06 PM, Jason Voorhees wrote:
> On Thu, Apr 28, 2011 at 1:43 PM, John Drescher wrote:
>> On Thu, Apr 28, 2011 at 2:38 PM, John Drescher wrote:
/dev/mapper/mpath0:
Timing buffered disk reads: 622 MB in 3.00 seconds = 207.20 MB/sec
>>> That is a raid. But
On Thu, Apr 28, 2011 at 1:43 PM, John Drescher wrote:
> On Thu, Apr 28, 2011 at 2:38 PM, John Drescher wrote:
>>> /dev/mapper/mpath0:
>>> Timing buffered disk reads: 622 MB in 3.00 seconds = 207.20 MB/sec
>>>
>> That is a raid. But you still may not be able to sustain over 100MB/s
>> of somewh
On Thu, Apr 28, 2011 at 2:38 PM, John Drescher wrote:
>> /dev/mapper/mpath0:
>> Timing buffered disk reads: 622 MB in 3.00 seconds = 207.20 MB/sec
>>
> That is a raid. But you still may not be able to sustain over 100MB/s
> of somewhat random reads. Remember that hdparm is only measuring
> sequ
> /dev/mapper/mpath0:
> Timing buffered disk reads: 622 MB in 3.00 seconds = 207.20 MB/sec
>
That is a raid. But you still may not be able to sustain over 100MB/s
of somewhat random reads. Remember that hdparm is only measuring
sequential performance of large reads.
--
John M. Drescher
-
On Thu, Apr 28, 2011 at 12:01 PM, John Drescher wrote:
>> So do you believe these speeds of my backups are normal? I though my
>> Library tape with LTO-5 tapes could write at 140 MB/s approx. It isn't
>> possible to achieve higher speeds?
>
> You need to speed up your source filesystem to achieve
> So do you believe these speeds of my backups are normal? I though my
> Library tape with LTO-5 tapes could write at 140 MB/s approx. It isn't
> possible to achieve higher speeds?
You need to speed up your source filesystem to achieve better
performance. Use raid10 or get a SSD. It has nothing at
On Thu, Apr 28, 2011 at 11:41 AM, John Drescher wrote:
>> How can I know where's the bottleneck? I'm using an ext4 filesystem.
>> Are these tests useful?
>>
>> [root@qsrpsbk1 ~]# hdparm -t /dev/sda
>>
>> /dev/sda:
>> Timing buffered disk reads: 370 MB in 3.01 seconds = 122.89 MB/sec
>> [root@qs
> How can I know where's the bottleneck? I'm using an ext4 filesystem.
> Are these tests useful?
>
> [root@qsrpsbk1 ~]# hdparm -t /dev/sda
>
> /dev/sda:
> Timing buffered disk reads: 370 MB in 3.01 seconds = 122.89 MB/sec
> [root@qsrpsbk1 ~]# hdparm -tT /dev/sda
>
> /dev/sda:
> Timing cached re
On Thu, Apr 28, 2011 at 10:30 AM, John Drescher wrote:
>> No, there are just a "normal" number of files from a shared folder of
>> my fileserver with spreadsheets, documents, images, PDFs, just
>> information of final users.
>>
>
> The performance problem is probably filesystem performance. A sing
> The performance problem is probably filesystem performance. A single
> hard drive will only hit 100 MB/s if you are baking up files that are
> a few hundred MB.
>
>
> --
> John M. Drescher
>
How could I run some tests to verify this? I'm running MySQL server in
the same host where Bacula is inst
Did you activated attribute spooling ( and maybe data spooling too if
you use LTO )?
2011/4/28 Jason Voorhees :
> Hi:
>
> On Thu, Apr 28, 2011 at 10:19 AM, John Drescher wrote:
>> On Thu, Apr 28, 2011 at 11:08 AM, Jason Voorhees
>> wrote:
>>> Hi:
>>>
>>> I'm running Bacula 5.0.3 in RHEL 6.0 x86
> No, there are just a "normal" number of files from a shared folder of
> my fileserver with spreadsheets, documents, images, PDFs, just
> information of final users.
>
The performance problem is probably filesystem performance. A single
hard drive will only hit 100 MB/s if you are baking up files
Try changing your Maximum Network Buffer size in your bacula-sd config.
Something like
Maximum Network Buffer Size = 262144 #65536
Maximum block size = 262144
Keep in mind that this will make your sd unable to read previous
backups, IIRC.
Search archives for this parameter, e.g.
http://old.
Hi:
On Thu, Apr 28, 2011 at 10:19 AM, John Drescher wrote:
> On Thu, Apr 28, 2011 at 11:08 AM, Jason Voorhees wrote:
>> Hi:
>>
>> I'm running Bacula 5.0.3 in RHEL 6.0 x86_64 with a Library tape IBM
>> TS3100 with hardware compression enabled and software (Bacula)
>> compression disabled, using L
On Thu, Apr 28, 2011 at 11:08 AM, Jason Voorhees wrote:
> Hi:
>
> I'm running Bacula 5.0.3 in RHEL 6.0 x86_64 with a Library tape IBM
> TS3100 with hardware compression enabled and software (Bacula)
> compression disabled, using LTO-5 tapes. I have a Gigabit Ethernet
> network and iperf tests repo
Hi:
I'm running Bacula 5.0.3 in RHEL 6.0 x86_64 with a Library tape IBM
TS3100 with hardware compression enabled and software (Bacula)
compression disabled, using LTO-5 tapes. I have a Gigabit Ethernet
network and iperf tests report me a bandwidth of 112 MB/s.
I'm not using any spooling configura
32 matches
Mail list logo