This post is more than 5 years old
4 Posts
0
1870
PowerConnect 5524 Fault Tolerant Setup
We have a Hyper-V 2012 R2 Cluster with several host nodes, which are all Dell PowerEdge Servers, connected to two PowerConnect 5524 switches, in a Master/Master-Backup setup, which are in turn connected to a MD3220i SAN.
The idea of this setup was that as long as only one piece of hardware failed then the cluster services would still successfully run.
This was working fine until the other day when the power went out for one of the switches and one of the power supplies for the SAN. This then led to a loss in connectivity between some of the cluster nodes and the volumes within the SAN.
The SAN did lose some connectivity but it remained powered on throughout so we don't believe this is the point of failure. Looking at the logs on the switches it appears as though there was a complete loss in connectivity for the Master. However this begs to question, is this truly a fault tolerant system.
Upon some investigation it appears as though some others, who have a similar setup to us, have found there is a loss of connectivity as the master-backup takes the place of the master switch. There has been mention of using an option called FastLink/PortFast which might speed up the process and thus make the likelihood of failure of connectivity lower.
Anyhow, does anyone know how the switches should be configured in order to ensure that our services are fault tolerant?
Thanks in advance to any replies.
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
1
February 20th, 2017 06:00
Some model switches have a non-stop forwarding feature that can help with network interruptions during a master failure. However the 5500 series does not offer this feature. When the master unit fails there will be a moment of interruption as the backup unit populates dynamic tables and initiates the running config.
The portfast feature allows an interface to enter the forwarding state right away, rather than going through the spanning tree learning process. This is typically used on edge facing ports, and can help interfaces come up quicker during times of topology change. the interface command is spanning-tree portfast.
Firmware can also help with overall operability. The firmware updates have included some stacking improvements. If the stack is currently not up to date, it would not hurt to update.
http://dell.to/1XMFWxN
DSSSysAdmin
4 Posts
0
February 17th, 2017 09:00
if the above question is a bit too broad then, what ports should I be using fastlink on?
Does FastLink make a noticeable difference?
Will Fastlink help to keep the cluster services to continue if the master switch goes down?
DSSSysAdmin
4 Posts
0
February 20th, 2017 06:00
Ok that's cleared up a few things.
When you mention "edge facing ports" is that referring to ports that are connected to non-networking devices, for example servers and SANs?
The firmware might require updating. Thank you for the link to the update.
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
February 20th, 2017 07:00
Correct, edge ports will basically be any VLAN unaware device. Including Workstations, most servers, printers, some firewalls, SANs.
DSSSysAdmin
4 Posts
0
February 21st, 2017 01:00
Nice.
Just two more questions:
Thank you very much for the help thus far.
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
1
February 21st, 2017 06:00
You should not see any connectivity loss by changing to portfast. The firmware download includes a PDF with step by step instructions on how to apply the firmware. These instructions are the recommended method with the least amount of downtime. The instructions also include a recovery option using XMODEM. The use of XMODEM will take a long time and should only be used as a last effort.