I have a unit that is experiencing some write bottle necks and back to back consistency points at times. I never see any of the disks busier than 60% even during the periods where I notice lots of CpfromCps.
The unit has 2 aggregates of 50 disk with a raid size of 18. 2 raid groups are full and the last only has 14 disks. Looking at a statit during normal operation The smaller raid group is not any busier than the larger rgs. (I'm going to try to get a statit while the back to back cps are occurring and see if this holds true)
During the periods of back to back cps I am seeing about 2500 writes ops and 7k read ops. This is a IBM rebranded N6240 with flexscale cards. (FAS3240)
So, the question is, I am thinking of enabling free_space_realloc on the aggregate and possibly read_realloc on the volumes. (sis is not in use on this unit)
Is there a way to measure the potential benefit of free_space_realloc, or rather how fragmented the aggregate is. I would not have expected it to be very fragmented at 60% used, but I would aslo expect to get maybe closer to 5k write ops before seeing back to back copies.
Questions, comments, warnings, random thoughts?
-JMS
For the fourth consecutive year, Independent Health is the highest ranked health insurance plan in the New York/New Jersey region by J.D. Power and Associates 2013 Member Health Plan StudySM.
For the sixth consecutive year, Independent Health was named one of the best companies to work for in New York State.
Go to www.independenthealth.com to view careers and available positions.
We are an Equal Opportunity Employer.
CONFIDENTIALITY NOTICE: This e-mail and its attachments (collectively referred to as "e-mail") may contain confidential information that is privileged and protected from disclosure by Federal and State confidentiality laws, rules or regulations. This e-mail is intended for the designated addressee only. If you are not the designated addressee, you are notified that any disclosure, copying or distribution of this e-mail may be unlawful and may subject you to legal consequences. If you have received this e-mail in error, please contact me immediately by telephone at (716) 631 - 3001 and delete the e-mail from your computer immediately. Thank you for your attention.
I'm new to this, so anyone feel free to correct me if i'm wrong, but as far as i saw... A good starting point to see if you could benefit from free space reallocation is to look at how many iops your controller is "wasting" trying to find free blocks to write a parity-wide chain:
take a look at those stats:
aggregate:*:user_writes aggregate:*:cp_reads
if the cp_reads are far above the user writes, it is likely you could benefit from a reallocation.
________________________________ Lorenzo Coppitelli Sinergy S.p.A.fre Viale Regina Margherita 269 00198 Roma (RM) ________________________________ Da: toasters-bounces@teaparty.net [toasters-bounces@teaparty.net] per conto di Jordan Slingerland [Jordan.Slingerland@independenthealth.com] Inviato: venerdì 6 dicembre 2013 20.32 A: Toasters@teaparty.net (Toasters@teaparty.net) Oggetto: free_space_realloc and measuring potentical benifit of free space optimization
I have a unit that is experiencing some write bottle necks and back to back consistency points at times. I never see any of the disks busier than 60% even during the periods where I notice lots of CpfromCps.
The unit has 2 aggregates of 50 disk with a raid size of 18. 2 raid groups are full and the last only has 14 disks. Looking at a statit during normal operation The smaller raid group is not any busier than the larger rgs. (I’m going to try to get a statit while the back to back cps are occurring and see if this holds true)
During the periods of back to back cps I am seeing about 2500 writes ops and 7k read ops. This is a IBM rebranded N6240 with flexscale cards. (FAS3240)
So, the question is, I am thinking of enabling free_space_realloc on the aggregate and possibly read_realloc on the volumes. (sis is not in use on this unit)
Is there a way to measure the potential benefit of free_space_realloc, or rather how fragmented the aggregate is. I would not have expected it to be very fragmented at 60% used, but I would aslo expect to get maybe closer to 5k write ops before seeing back to back copies.
Questions, comments, warnings, random thoughts?
-JMS
For the fourth consecutive year, Independent Health is the highest ranked health insurance plan in the New York/New Jersey region by J.D. Power and Associates 2013 Member Health Plan StudySM.
For the sixth consecutive year, Independent Health was named one of the best companies to work for in New York State.
Go to www.independenthealth.com to view careers and available positions.
We are an Equal Opportunity Employer.
CONFIDENTIALITY NOTICE: This e-mail and its attachments (collectively referred to as "e-mail") may contain confidential information that is privileged and protected from disclosure by Federal and State confidentiality laws, rules or regulations. This e-mail is intended for the designated addressee only. If you are not the designated addressee, you are notified that any disclosure, copying or distribution of this e-mail may be unlawful and may subject you to legal consequences. If you have received this e-mail in error, please contact me immediately by telephone at (716) 631 - 3001 and delete the e-mail from your computer immediately. Thank you for your attention.
I'd say more than .5. 1:1 is far too late.
Sent from my iPhone
On Dec 9, 2013, at 2:02 AM, Coppitelli Lorenzo L.Coppitelli@sinergy.it wrote:
I'm new to this, so anyone feel free to correct me if i'm wrong, but as far as i saw... A good starting point to see if you could benefit from free space reallocation is to look at how many iops your controller is "wasting" trying to find free blocks to write a parity-wide chain:
take a look at those stats:
aggregate:*:user_writes aggregate:*:cp_reads
if the cp_reads are far above the user writes, it is likely you could benefit from a reallocation.
Lorenzo Coppitelli Sinergy S.p.A.fre Viale Regina Margherita 269 00198 Roma (RM) Da: toasters-bounces@teaparty.net [toasters-bounces@teaparty.net] per conto di Jordan Slingerland [Jordan.Slingerland@independenthealth.com] Inviato: venerdì 6 dicembre 2013 20.32 A: Toasters@teaparty.net (Toasters@teaparty.net) Oggetto: free_space_realloc and measuring potentical benifit of free space optimization
I have a unit that is experiencing some write bottle necks and back to back consistency points at times. I never see any of the disks busier than 60% even during the periods where I notice lots of CpfromCps.
The unit has 2 aggregates of 50 disk with a raid size of 18. 2 raid groups are full and the last only has 14 disks. Looking at a statit during normal operation The smaller raid group is not any busier than the larger rgs. (I’m going to try to get a statit while the back to back cps are occurring and see if this holds true)
During the periods of back to back cps I am seeing about 2500 writes ops and 7k read ops. This is a IBM rebranded N6240 with flexscale cards. (FAS3240)
So, the question is, I am thinking of enabling free_space_realloc on the aggregate and possibly read_realloc on the volumes. (sis is not in use on this unit)
Is there a way to measure the potential benefit of free_space_realloc, or rather how fragmented the aggregate is. I would not have expected it to be very fragmented at 60% used, but I would aslo expect to get maybe closer to 5k write ops before seeing back to back copies.
Questions, comments, warnings, random thoughts?
-JMS
For the fourth consecutive year, Independent Health is the highest ranked health insurance plan in the New York/New Jersey region by J.D. Power and Associates 2013 Member Health Plan StudySM.
For the sixth consecutive year, Independent Health was named one of the best companies to work for in New York State.
Go to www.independenthealth.com to view careers and available positions.
We are an Equal Opportunity Employer.
CONFIDENTIALITY NOTICE: This e-mail and its attachments (collectively referred to as "e-mail") may contain confidential information that is privileged and protected from disclosure by Federal and State confidentiality laws, rules or regulations. This e-mail is intended for the designated addressee only. If you are not the designated addressee, you are notified that any disclosure, copying or distribution of this e-mail may be unlawful and may subject you to legal consequences. If you have received this e-mail in error, please contact me immediately by telephone at (716) 631 - 3001 and delete the e-mail from your computer immediately. Thank you for your attention.
Toasters mailing list Toasters@teaparty.net http://www.teaparty.net/mailman/listinfo/toasters
-
Coppitelli Lorenzo -
Jeff Mohler -
Jordan Slingerland