|
|
|
Forum Newbie
Group: Forum Members
Last Login: Tuesday, November 27, 2007 11:46 AM
Posts: 3,
Visits: 9
|
|
Running two VSA's, one each on Dell servers with internal raid one. I had no problems creating the volumes or presenting them to both hosts. We have verified networking is good, not bottle necks in cpu, memory, i/o.
After the formats complete, copying a 7Mb file to the Lefthand volume take 20 minutes.
RJL
|
|
|
|
|
Supreme Being
Group: Moderators
Last Login: Yesterday @ 2:22 PM
Posts: 101,
Visits: 354
|
|
| Can you give more detail? Correctly configured VSA should be able to do 50mb/s write so something is definitely not sane about your configuration. But where to begin. I'll accept that it is not a network issue (for now) because even if you were only running on fast ethernet it should be able to do 7MB in one second. Taking that long to copy that small of a file smells like there is some contention over the volume or a resource the VSA relies on. Do either VSA live on a datastore with other VMs using the same disk directly? Do the VSA have their CPU and memory reserved? 2000 and 1024 respectively? Where is the file being copied from? Is the VMkernel for iSCSI on the same switch as the VSA? How many HDD, and of what type are in the config? I'd assume 2 in each server since you said R1. I'd like to help get you running close to our expectations.
Adam C
Product Manager
LeftHand Networks
|
|
|
|
|
Forum Newbie
Group: Forum Members
Last Login: Tuesday, November 27, 2007 11:46 AM
Posts: 3,
Visits: 9
|
|
Do either VSA live on a datastore with other VMs using the same disk directly?
No. Each VSA lives on an internal datastore. No other VM's reside on internal disk.
Do the VSA have their CPU and memory reserved? 2000 and 1024 respectively?
Yes they are.
Where is the file being copied from?
I copied files from /var/log/vmkernel.x
Is the VMkernel for iSCSI on the same switch as the VSA?
Yes they are. These servers only have two nics and are bonded.
How many HDD, and of what type are in the config? I'd assume 2 in each server since you said R1.
There are two 73GB drives per server in a raid 1 with the Dell Perc controllers.
Note, even a vdf -h command hangs for 20 seconds when it hits the Lefthand datastore.
|
|
|
|
|
Supreme Being
Group: Moderators
Last Login: Yesterday @ 2:22 PM
Posts: 101,
Visits: 354
|
|
| Sounds like you are following most best practices. Still something is definitely not right though. I'd not expect VSA built out of only 2 mirrored drives to provide much in IOPS, couple hundred at best, but we do run performance tests on blades using just two mirrored drives and they should function just fine. I quickly tried vdf -h on my vsa and got an immediate response. can you do a vmkping from the esx server to the virtual ip address of the storage cluster? Just to make sure there is not high latency there for some reason.
Adam C
Product Manager
LeftHand Networks
|
|
|
|
|
Forum Newbie
Group: Forum Members
Last Login: Tuesday, November 27, 2007 11:46 AM
Posts: 3,
Visits: 9
|
|
Here is the vmkping output
[root@vmpesx1 sbin]# ./vmkping 192.168.0.83
PING 192.168.0.83 (192.168.0.83): 56 data bytes
64 bytes from 192.168.0.83: icmp_seq=0 ttl=64 time=0.672 ms
64 bytes from 192.168.0.83: icmp_seq=1 ttl=64 time=0.507 ms
64 bytes from 192.168.0.83: icmp_seq=2 ttl=64 time=0.516 ms
--- 192.168.0.83 ping statistics ---
3 packets transmitted, 3 packets received, 0% packet loss
round-trip min/avg/max = 0.507/0.565/0.672 ms
|
|
|
|