Unsolved
This post is more than 5 years old
63 Posts
0
245244
Best pratice for vSphere 5
Hi all,
Considering of the following enviroment:
1) 4 Physical Dell R710 Servers,each server has 16 Nics (Gigabit) will install
vSphere 5 std.
2) 2 Switches with stack function for SAN Storage
3) 2 Dell Equallogic PS4000xv SAN (Dual controller )
4) 2 Switches for virtual machine traffic
Regarding to the networking, I plan do create some vSwitches on each physical
server as follows
1. vSwitch0 - used for iSCSI storage
6 NICs Teamed, with IP-hash teaming policy, multipathing with iSCSI
Storage;and the stroage load balancing is Round Rodin(vmware)
( vmware suggests use 2
NICs for 1 IP storage traget, I am not sure)
2. vSwitch1 - used for virtual machine
6 NICs Teamed for virtual machine traffic, with IP-hash policy
3. vSwitch2 - for managment
2 NICs Teamed
4. vSwitch3 - vMotion
2 NICs teamed
Would you like kindly give me some suggestions?
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
May 21st, 2012 08:00
It's hard to monitor it long time since the connections get balanced and that resets the counters.
The best way is install the MEM module, since that knows what members have the data being requested or the member the right needs to go to. So the connections will always be balanced correctly.
You could run ESXTOP and output to a file and monitor that way. There's a util to graph that output.
In the Performance tab you can monitor the NICs there as well.
Regards,
sketchy00
203 Posts
0
May 21st, 2012 08:00
Don, what have you found as the most effective way to verify that both paths are being balanced correctly?
DavidSte
13 Posts
0
April 19th, 2013 06:00
I'm just wondering what the reasoning behind disabling LRO is - can anyone confirm the reasoning behind why it needs to be disabled?
Thanks, David
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
April 19th, 2013 08:00
It works in a similar way to DelayedACK. LRO holds onto data until a buffer is filled then sends it. This host sees this as artficial latency. Since it "sent" the data but no ACK has been receieved. iSCSI is very sensitive to latency.
Regards,
Damien Calvert
1 Rookie
1 Rookie
•
16 Posts
0
May 29th, 2013 00:00
Thanks for all this fantastic information.
I've just upgrade to ESX 5.1 and so If i'm installing the Equallogic Multipathing plugin, then all I also need to do is disable LRO and Delayed ACK on the iscsi adapters?
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
May 29th, 2013 09:00
Those are the three big ones. Adding additional virtual SCSI adapters in your VMs, when you havd multiple VMDKs or Raw Device Mapped drives is another. (That's discussed earlier in this thread). Not putting too many high I/O VMs on the same volume is also good. So more smaller volumes tend to work better than one or two MEGA volumes.
Also make sure you have updated to the latest build of 5.1.
Regards,
austin-t
84 Posts
0
June 7th, 2013 10:00
Hi Don,
Can you explain why LSO doesn't need disabling if LRO does please?
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
June 7th, 2013 11:00
Hello,
re: LSO. I've not seen that hold onto packets, thus delaying ACKs and inducing latency into the process.
Don
bealdrid
56 Posts
0
June 8th, 2013 15:00
Hi Don, just curious, can we expect an updated official vSphere with EqualLogic best practices document/guide from EqualLogic anytime soon?
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
June 8th, 2013 15:00
Hello,
There is one already in the works.
Regards,