Disk Enclosures
1745784 Members
3984 Online
108722 Solutions
New Discussion юеВ

Prevent raid 5 on AutoRaid

 
Terrence Johnson
Occasional Contributor

Prevent raid 5 on AutoRaid

I'm running a Progress application on my K570 and my database is stored on an AutoRaid. Since raid 5 has the effect of killing the performance on a Progress app, I want to prevent keep the disk usage from going over 50%. But 50% of what? Do I look at usage at the volume group level, the filesystem level, or the entire array? And then, how much un-allocated physical disk should I keep?
...just round up the usual suspects
10 REPLIES 10
Patrick Wallek
Honored Contributor

Re: Prevent raid 5 on AutoRaid

To keep the data in an Autoraid arrya from being migrated to Raid5, you should only allocate 50% of the space to VGs and LVs on your system. This will ensure that there is always enough space available to keep everything in Raid 0+1.
A. Clay Stephenson
Acclaimed Contributor

Re: Prevent raid 5 on AutoRaid

The AutoRAID knows nothing about filesystems, volume groups, or logical volumes. All it knows is LUNS. Keep the total space of the array allocated as LUN's at around 50% or less and the AutoRAID will stay in RAID 1/0 at all times. The other big key to performance is to configure VG's as two identically sized LUN's. LUN0's primary path should be through Controller X (alternate Y) and LUN1's primary PATH should be through Controller Y (alternate X). Stripe each LVOL in this VG across both LUN's in 64K chunks.
If it ain't broke, I can fix that.
S.K. Chan
Honored Contributor

Re: Prevent raid 5 on AutoRaid

50% or less of the total LUN space and it'll not go to raid5. You would run ..
# arraydsp -a |more
and look at the section "Disk space Usage" to calculate your usage.
Terrence Johnson
Occasional Contributor

Re: Prevent raid 5 on AutoRaid

Thank you all for your answers. Now when I look at the results of 'arraydsp -a' I see that Total Physical = 88962 mb, Allocated to LUNS = 43808 mb, Active Hot Spare = 17366 mb, Used for Redundancy = 23285 and Unallocated (avail for LUNS) = 4503. This would lead me to believe that I'm not in any danger of migrating to RAID 5, and that 'bdf' is not a good indicator of "disk" space usage, just filesystem usage, no?
...just round up the usual suspects
S.K. Chan
Honored Contributor

Re: Prevent raid 5 on AutoRaid

You're good .. 43808x2 has not exceed the "total physical" yet. You may want to take a look more into the "arraydsp" command that would display some performance stats. For instance ..
# arraydsp -m 03100800 03101500
and take a look at "Relocate Block", that number indicates how much data is moved between raid0/1 and raid5, obviously yours would be zero. It's one way to gauge the array performance. If that value is big, performance goes down. Man pages of arraydsp will explain more (I think).
Eugeny Brychkov
Honored Contributor

Re: Prevent raid 5 on AutoRaid

Terrence,
would you please attach 'arraydsp -a' and 'logprint' outputs zipped to your next reply?
Eugeny
Terrence Johnson
Occasional Contributor

Re: Prevent raid 5 on AutoRaid

Eugeny,

...that's another problem that I just finished talking with HP Support about. Our ARMServer needs patching and our AutoRaid firmware needs updating to get logprint and arraydsp to function properly. I'm getting to the point where I've almost forgotten the original problem. <:|
...just round up the usual suspects
Eugeny Brychkov
Honored Contributor

Re: Prevent raid 5 on AutoRaid

Did you think if you will solve this software/patching/firmware issue perfomance issue will go away? Do a system backup, and install recent Diags, ARMmanager (not sure if it's included with patches), GR and HWE patch bundles from CDs of one revision. Then update (ask HP to update) AutoRAID's firmware. Then look to its diagnostic logs (arraydsp -a, logprint) to make sure hardware is running ok and only then measure perfomance.
You know it's useless to do perfomance troubleshooting on 'non-optimal' system.
Good luck!
Eugeny
Roger_22
Trusted Contributor

Re: Prevent raid 5 on AutoRaid

Any progress? Once you have the new firmware installed, you need to verify the settings are ???Normal??? Resilience, Prefetch Enabled, and Optimize Enabled.

HP18, and these settings, can make a dramatic difference. Watch your application performance for a while, if it doesn???t get any better then consider RAID 1+0. I???ve also seen additional write cache help, you might consider that also.

If you reply with the VA model, and the number and size of the disks, I can compute the magic capacity for the RGs ??? that is the LUN capacity at which the RG will be RAID 1+0 only. It???s not exactly 50% of anything???.