Subscribe

Lots of CP Type H on filers and possible performance issues

I have 4 filers and they are almost all exclusively showing CP type of H.  I understand that it is a high watermark CP, but what no one has been able to explain to me is, what are the affects on performance when these are essentially the only CP types I am seeing on my system.  I never see a full CP.

The filers are all FAS3240 and most disks are 600GB 15K SAS in DS4243 shelves.  I do have 2 shelves of 1TB BSAS.  The intervals on both sysstats are 1 second.  Sorry for the long post, but I wanted to provide a good set of data.

What most concerns me is that my filers appear to be running near their limits and I typically only push upwards of 11k IOPS.  I was under the impression the FAS3240 was capable of close to 40k IOPS...

I am attaching a sysstat during a typical time throughout the day and one during a more intense workload (DB2 processing).

Thanks in advance!

Don

Re: Lots of CP Type H on filers and possible performance issues

My Filer (3140 DoT:8.0.2) is experiencing the same same issue. Did you have any luck finding an answer to this?

Re: Lots of CP Type H on filers and possible performance issues

Please post any info as seeing the same CP H and slow performance

Re: Lots of CP Type H on filers and possible performance issues

The following KB article provides a brief explanation of the "H" type of CP:

https://kb.netapp.com/support/index?page=content&id=3012539

This would be a good question to post in the NetApp Support Community, however if your issue is urgent please open a case with NetApp Technical Support.

Thanks,

Christine

Re: Lots of CP Type H on filers and possible performance issues

Hi,

Lots of writing to the disks or lots of modified data?

H = number of buffers holding data that has been modified but not yet written

to disk exceeds a threshold, then a CP from high watermark is triggered

I can see, that you got some "B" letters as well, those are bad, most of the time it means,

there are too many write requestes and filer can't hadle those. NVMEM is full, but data haven't been written to the disks yet.

This will be showing a high latency issues for end user.

Intense sysstat also shows, that CPU is nearly 100%

I think command "statit" could be use in this case. (Can't remeber exact syntax at the moment)

Br.

Ismo