Raid1: sdb has a lot mor work then sda
Hello,
at the moment I'm trying to find a bottleneck on my lamp-server with opensuse 11.4 (Kernel 2.6.37.6)
Sometimes the system has a very poor performance because of a high io-wait. If I watch the systems disk-access with "atop -dD" I can see that sda is the most of the time at a load of 10%, sdb sometimes is at 100% or higher at the same time.
In my opinion in a Raid1-System both disk should have nearly the same load.
Or may I wrong whis this?
Both harddisks where changed 3 Weeks ago, /proc/mdstat shows that the rebuild was successfull and the array is functional. Today I've changed th SATA-Cable and the Port on the Maindboard of sdb, but the behaviour is still the same. Both disks passed the extended smart-self-test.
Any ideas about that?
Regards
Daniel
Signature · 9 lines
Daniel Spannbauer Software Entwicklung marco Systemanalyse und Entwicklung GmbH Tel +49 8333 9233-27 Fax -11 Rechbergstr. 4 - 6, D 87727 Babenhausen Mobil +49 171 4033220 http://www.marco.de/ Email [email protected] Geschäftsführer Martin Reuter HRB 171775 Amtsgericht München -- To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
Hello,
here is a screendump of a atop -dD in such a situation.
Regards
Daniel
Am 04/24/2012 05:00 PM, schrieb Roberto Spadim:
Quoted 62 lines
it´s a read, a write, or both operation?
about load i think it´s a very big read (sdb) with a very low write (sda) the disk balance on raid1 is based on nearest (current position - next read position) disk
on a sequencial read only one disk is better usedi checked that raid1 'have this problem' but it´s a hardware limit, i didn´t found a solution, maybe others friends can found
Em 24 de abril de 2012 11:46, Daniel Spannbauer <[email protected] <mailto:[email protected]>> escreveu:
Hello,
at the moment I'm trying to find a bottleneck on my lamp-server with opensuse 11.4 (Kernel 2.6.37.6)
Sometimes the system has a very poor performance because of a high io-wait.
If I watch the systems disk-access with "atop -dD" I can see that sda is the most of the time at a load of 10%, sdb sometimes is at 100% or higher at the same time.In my opinion in a Raid1-System both disk should have nearly the same load.
Or may I wrong whis this?
Both harddisks where changed 3 Weeks ago,
/proc/mdstatshows that the rebuild was successfull and the array is functional. Today I've changed th SATA-Cable and the Port on the Maindboard of sdb, but the behaviour is still the same. Both disks passed the extended smart-self-test.Any ideas about that?
Regards
Daniel
--
Daniel Spannbauer Software Entwicklung
marco Systemanalyse und Entwicklung GmbH Tel +49 8333 9233-27
<tel:%2B49%208333%209233-27> Fax -11
Rechbergstr. 4 - 6, D 87727 Babenhausen Mobil +49 171 4033220
<tel:%2B49%20171%204033220>
http://www.marco.de/ Email [email protected]
<mailto:[email protected]>
Geschäftsführer Martin Reuter HRB 171775 Amtsgericht München
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to [email protected] <mailto:[email protected]>
More majordomo info at http://vger.kernel.org/majordomo-info.htmlSignature · 2 lines
Roberto Spadim Spadim Technology / SPAEmpresarial
Signature · 9 lines
Daniel Spannbauer Software Entwicklung marco Systemanalyse und Entwicklung GmbH Tel +49 8333 9233-27 Fax -11 Rechbergstr. 4 - 6, D 87727 Babenhausen Mobil +49 171 4033220 http://www.marco.de/ Email [email protected] Geschäftsführer Martin Reuter HRB 171775 Amtsgericht München -- To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
it´s a read, a write, or both operation?
about load i think it´s a very big read (sdb) with a very low write (sda) the disk balance on raid1 is based on nearest (current position - next read position) disk
on a sequencial read only one disk is better used
i checked that raid1 'have this problem' but it´s a hardware limit, i didn´t found a solution yet , maybe others friends can found i made some read balance method some time ago, and the results is constant 1% of performace improvement on mixed read/write scenario, but i didn´t saw it as a good performace improvement since it must have many parameters to configure (time of disk seeks), and the load isn´t proportional, sometimes a disk use is very bigger than the other disk
saw raid1 as a parallel/secure work, not as a performace tweak
--
Roberto Spadim
Spadim Technology / SPAEmpresarial
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
ops, sorry may email sent two messages
Signature · 2 lines
Roberto Spadim Spadim Technology / SPAEmpresarial
Quoted 19 lines · Daniel Spannbauer
Hello,
at the moment I'm trying to find a bottleneck on my lamp-server with opensuse 11.4 (Kernel 2.6.37.6)
Sometimes the system has a very poor performance because of a high io-wait. If I watch the systems disk-access with "atop -dD" I can see that sda is the most of the time at a load of 10%, sdb sometimes is at 100% or higher at the same time.
In my opinion in a Raid1-System both disk should have nearly the same load.
Or may I wrong whis this?
Both harddisks where changed 3 Weeks ago,
/proc/mdstatshows that the rebuild was successfull and the array is functional. Today I've changed th SATA-Cable and the Port on the Maindboard of sdb, but the behaviour is still the same. Both disks passed the extended smart-self-test.
Quoted 5 lines
Any ideas about that?
Regards
Daniel
Are the disks more or less the same in all SMART attributes (from smartctl -a)?
I have seen such a behaviour in the case of one disk being marginal, and/or having a couple of bad spots which need several trials to be read. This always also resulted in longer "smartctl -t long" times than usual, but did not always show up clearly in the SMART output (or I looked at the wrong attributes).
I'd try to exchange the disk.
HTH,
Kay