E5725 I/O Problem | NetApp | Page 1

fresh trout Jul 13, 2023, 11:57 AM

#

E5725 I/O Problem

upper quest Jul 13, 2023, 9:37 PM

#

How many drives are present in the system? And can you elaborate on the raid configuration? Especially raid group size?
Also is the Huawei connected via the same HBA on the VMware side? Read/write over the same HBA significantly decreases storage vmotion performance.

fresh trout Jul 14, 2023, 6:04 AM

#

upper quest How many drives are present in the system? And can you elaborate on the raid con...

Total 72 drives (1.8tb sas * 24e/a)
Internal 24e/a
De224 #1 24e/a
De224 #2 24e/a

We configured RAID-6 and we used shelf1 disks (shelf1-00~23)

And I think HBA will be different because it is zoned to SAN SW, not DAS configuration.

#

I m so confused!

upper quest Jul 14, 2023, 6:14 AM

#

A sas drive can deliver around 200 Io/sec max. With raid and hot spare, you have 20 disks processing data.
If ESXi reads over the same HBA as it's writes, that is probably it.

#

You can get much more performance by using a dedicated hba in Esxi for the E-series and using DDP over all drives with multiple volumes as VMFS datastores.
Also analyze if the reading from Huawei is not a bottle neck

fresh trout Jul 14, 2023, 6:19 AM

#

If the volume is configured with DDP, the available capacity is reduced, so we inevitably configured with Raid6. Will it migrate faster if I configure it with DAS?

upper quest Jul 14, 2023, 11:01 AM

#

I hate to say it, but it depends where the bottleneck is.

hybrid fog Jul 14, 2023, 12:18 PM

#

those 1.5k IOPS are at which block size? 1.5k IOPS at 1k are abyssimal, while 1.5K IOPS at 1meg are quite okay, for example

upper quest Jul 14, 2023, 2:53 PM

#

Assuming a supported VM environment, the blocksize of VMFS-6 is 1 MB

#E5725 I/O Problem