• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Cisco switch config for optimal VM performance

T

Thuleman

Supreme [H]ardness
Joined
Apr 13, 2004
Messages
5,833
So I have been in touch with the network folks as I am getting fairly frequent

HA agent has an error : Time-out while communicating with the HA agent error 10/13/2009 4:04:05 AM

errors. There's no activity on my servers during that time, so something else is happening to that switch, and they are looking into what that may be. However, as part of the discussion we are trying to figure out how to properly configure the switch to best serve my VMs and storage. Mind you, there's no substantial load on the systems, BUT I do expect that I will start to make a lot more use of vMotion soon(tm).

With that in mind, my production network and my private network are on the same switch. The "storage network", such as it is, is its own VLAN, but it shares it with very low private traffic between VMs.

Here's what the IT folks had to say about the switch:

1. What's the manufacturer and model number for this switch, which firmware is it running?

Cisco WS-C2960G-48TC-L running Cisco IOS Software, C2960 Software (C2960-LANBASE-M), Version 12.2(25)SEE1, RELEASE SOFTWARE (fc1)

2. Is each port rated for GigE? If not, see #3.

Yes, all ports are 10/100/1000 ports with speed/duplex set for auto/auto. They are all autonegotiated to 1000/full with the exception of your ILO ports which negotiated to 100/full. The switch has a 32gig backplane, so it would not be able to handle all 48 ports passing a full gig of traffic at the same time. The likelihood of this, however, is relatively small (currently your systems are the only ones connected to this switch).

3. Is each port group rated for GigE.

N/A

4. What cache amount is dedicated to each port?

Not certain. Here is the spec sheet for the switch series though:
http://www.cisco.com/en/US/prod/col...s6406/product_data_sheet0900aecd80322c0c.html

5. Is Flow Control set to ON (or Auto Negotiate for Cisco products)?

Input flow-control is set to off by default, but can be set to off/on/desired. Output flow control is not supported.

6. Is Symmetric Flow Control set to send and receive?

See 5

7. Is Asymmetric Flow Control set to receive?

See 5

8. Is Unicast Storm Control disabled?

Unicast storm-control is disabled.

9. Are Jumbo Frames supported? If so, are the ports configured for JF?

Jumbo Frames are supported. They have not been enabled on this switch. Doing so would require a scheduled outage to reload the switch.
Click to expand...

My actual question is, do any of these settings need to be changed. Is this switch adequate or should I be asking for a different make/model?
 
None of those settings would need to be changed, but its always good practice to hard set speed/duplex. Pretty much the standard for all Cisco switching. If there is not much activity going on between/from the hosts I HIGHLY doubt the switch is the culprit here. On everyvirtualization install ive done with Cisco switching(about 300+ SVI's on a few of them) ive never seen any issue as well as my multi-host lab at home.

I would bet a case that its a problem with your hosts but for troubleshooting sake, did you check the CPU history for the switch? Check out the SVI's and host facing interfaces for errors? Are you debugging at all, and if not do a debug level. Did you check the buffers? CAM space?

Also, dmesg on the hosts? anything abnormal?
 
xphil3 said:
None of those settings would need to be changed, but its always good practice to hard set speed/duplex.
Click to expand...

I'd agree in most situations, however we recently had a VMware health check done in our environment. The report they give you afterwards actually recommends maintaining autonegotiation: "Incorrect network speed and duplex settings can severly impace performance. Gigabit networks should be set to autonegotiate for optimum performance".
 
IMO, they just want you to autonegotiate because they don't want you to configure something manually incorrectly and then run into duplex mismatch problems and the like.
 
and because auto-neg can figure out the difference between crossover/etc cables.
 
lopoetve said:
and because auto-neg can figure out the difference between crossover/etc cables.
Click to expand...

EDIT - Ugh, never mind, I'm wrong. Autonegotiation is not required on gigabit interfaces that support Forced Mode Auto-MDIX (per a wikipedia article with crap for citation), but is required for MDIX to work otherwise.

Coulda sworn I remembered linking 2 interfaces that I'd hard set without a crossover... :confused:
 
Last edited:
You forgot 'port fast' on the switch ports but that is probably not the issue.

I would be more interested in the network setup of the actual ESX hosts, and if you have dedicated NICs for VM network, console/kernel, and vMotion.

Lastly, make sure your ESX hosts are synched to an NTP time source. It sounds crazy but I have seen this as the source for a lot of different time-out issues in ESX.

Check your port counters for dropped packets and pause frames (sh int, sh flow control).
 
You must log in or register to reply here.
Back
Top