System Administration
cancel
Showing results for 
Search instead for 
Did you mean: 

slow filesystem response suddenly

Carlos_172
Frequent Advisor

slow filesystem response suddenly

Hello,everybody.

We have a rp3440 hpux 11.11 two internal mirrored disk, attached to msa 1000. We have an oracle database 10G system.
I noticed sometimes, when system id doing rman backup datase to another hpux sytem through NFS filesystem, suddenly system drecrese perfomance dramatically. When I do "sar -d 2 100" , reponse time and service time of ALL DISKS is extremely poor, compare to normal activiti system.
Then I have to reboot system. I believe is something related to NFS. Last week I tried to copy one archive to another system trough NFS and I get same result.

Any suggestion??. Thank to all.
4 REPLIES
Johnson Punniyalingam
Honored Contributor

Re: slow filesystem response suddenly

Carlos_172
Frequent Advisor

Re: slow filesystem response suddenly

It's something strange that ALL others filesystems is hanging, It seems than an NFS operation cause than all filesystems hanging
Johnson Punniyalingam
Honored Contributor

Re: slow filesystem response suddenly

it's something strange that ALL others filesystems is hanging<<<

In this case, I will dig bit deeper , to clear doubts on the (disks) both Internal & external
any disk bottle necks .? Bad disk falling .? .etc something like that.

take help from (dmesg,vgdisplay,dd) commands to confirm or (cstm) to all disks are healthy.

pertaining to NFS "patch level also needs to be cross checked"
Problems are common to all, but attitude makes the difference
Hein van den Heuvel
Honored Contributor

Re: slow filesystem response suddenly

Carlos,

You mention response times, but not throughput.

Perhaps you maxed out MB/sec or IO/sec for the connection or the controller?

The MSA is probably connected with a single 1 gb (thick) fibre correct? Are you seeing 100 MB/sec?

Is the MSA cache battery good, adn the write caches active? How much physical cache memory in the MSA and how many LUNs presented out (cache is divided out between luns)

Can you please number to describe the problem rather than emotions (extremely poor)?
2 ms --> 20 ms? 200 ms?

" two internal mirrored disk, attached to msa 1000. "
Are those two internal disks, perhaps as system drives, in addition to the drives presented on the MSA?

How many disk are involved in the IOs testsed? Large (wide) raid sets, or simple mirrors? 10 krpm drives?

Hope this questions help some!
Regards,
Hein van den Heuvel
HvdH Performance Consulting