Fibre Channel (SAN)

Reply
dmj
New Contributor
dmj
Posts: 4
Registered: ‎11-01-2012

DCX failover

Hi.

I'm preparing two DCX 8510-4 with FabOS 7.0.1a for production use and I don't think I understand HA feature.

Every DCX goes with two CP.

CP0 has external IP a.b.c.d

CP1 - e.f.g.h

Plus, there's chassis IP: i.j.k.l

HA feature is enabled, so when I open i.j.k.l in web browser or by SSH, i'm logging into currently active CP. Standby CP can't do much (can't takeover HA and become Active, for example). hafailover or hareboot issued on active CP work just fine. But in the case of lost mgmt connectivity failover doesn't happen at all. So, say, CP0 is active, and his management connection goes down. Chassis IP i.j.k.l becomes unavailable, CP0 IP a.b.c.d becomes unavailable, CP1 is available via SSH as e.f.g.h but can't become active. The only workaround is to login via SSH onto CP1 as root and then ssh from CP1 to CP0 through secondary interface (which doesn't have anything to do with IP shown by ipaddrshow - in default configuration it's 10.0.0.5/10.0.0.6, while in ifconfig it's really 127.1.1.5/127.1.1.6).

The question is, can I force hafailover in case of link loss on management interface of active CP?

Moderator
Antonio Bongiorno TechHelp24
Posts: 3,804
Registered: ‎02-23-2004

Re: DCX failover

--->>>The question is, can I force hafailover in case of link loss on management interface of active CP?

theoretically, if a link is loss to management - both CP0 and CP1 -

( I've never see one Brocade Director that loss BOTH CP at one Time and this is very rare ) you cannot do nothing except trough the Serial Console, however if ha is enable "haenable" and one CP failed the standby CP become as active.

for details about "ha" refer the Command Reference Manuals

TechHelp24
dmj
New Contributor
dmj
Posts: 4
Registered: ‎11-01-2012

Re: DCX failover

I guess my english is too rusty. What I mean is not link loss on both CPs, the question is about active CP management link loss only, which leads to unavailability of chassis' IP as well. Standby CP is still available in this case, but failover doesn't happen automatically.

Moderator
Antonio Bongiorno TechHelp24
Posts: 3,804
Registered: ‎02-23-2004

Re: DCX failover

Is you question related about obvously scenario or is this the current situation ?

"hashow" give the output if ha is enable

if is not, issuing "haenable"

DCX use 3 IP address, TWO for CP and ONE for the Chassis, see example below.

usually you issuing the login trough the chassis IP and not to CP, now you are automatically logged into Active CP

if Active CP failed Standby become as Active

 

Chassis

Ethernet IP Address: 10.33.60.85

Ethernet Subnetmask: 255.255.240.0

CP0

Ethernet IP Address: 10.33.60.86

Ethernet Subnetmask: 255.255.240.0

Host Name: cp0

Gateway IP Address: 10.33.48.1

CP1

Ethernet IP Address: 10.33.60.87

Ethernet Subnetmask: 255.255.240.0

Host Name: cp1

Gateway IP Address: 10.33.48.1

TechHelp24
dmj
New Contributor
dmj
Posts: 4
Registered: ‎11-01-2012

Re: DCX failover

This question is related to possible scenario I'm concerned about.

Current config looks like this:

> ipaddrshow

CHASSIS

Ethernet IP Address: 10.10.0.250

Ethernet Subnetmask: 255.255.255.0

CP0

Ethernet IP Address: 10.10.0.235

Ethernet Subnetmask: 255.255.255.0

Host Name: cp0

Gateway IP Address: 10.10.0.254

CP1

Ethernet IP Address: 10.10.0.236

Ethernet Subnetmask: 255.255.255.0

Host Name: cp1

Gateway IP Address: 10.10.0.254

Backplane IP address of CP0 : 10.0.0.5

Backplane IP address of CP1 : 10.0.0.6

IPv6 Autoconfiguration Enabled: Yes

Local IPv6 Addresses:

IPv6 Gateways:

> hashow

Local CP (Slot 4, CP0): Active, Warm Recovered

Remote CP (Slot 5, CP1): Standby, Healthy

HA enabled, Heartbeat Up, HA State synchronized

I always connect to 10.10.0.250. I guess if CP0 fails completely, there will be failover (I don't know how to test this case though - by physically removing active CP blade, maybe?). But when I bring down the link on management interface of CP0, failover doesn't happen. So, for example, when CP0 is active and CP1 is standby:

I log into cisco mgmt switch and shut down interface corresponding to CP0.

10.10.0.235 becomes unavailable

10.10.0.250 becomes unavailable as well

10.10.0.236 is available, but failover doesn't happen automatically.

The only thing I can do in this case, is:

ssh root@10.10.0.236

(from CP1)> ssh 127.1.1.5 (10.0.0.5 doesn't work and there is no such address in ifconfig but it has 127.1.1.X on eth1)

(from CP0)> hafailover

Then CP1 becomes active and only then 10.10.0.250 becomes available.

The point is, if there is no failover in case of link loss on active CP, the whole concept of single chassis IP is kinda pointless. Am I doing something wrong here?

Valued Contributor
felipon
Posts: 670
Registered: ‎06-11-2010

Re: DCX failover

HI there,

the HA is designed to maintain High Availavility of the FC services, the switch does not consider the loss of the MGMT interface important enough (and neither do I) to force a failover. Bear in mind, that if the switch failed over everytime the lan link bounces, that would be very dangerous for the environment.

As you pointend out, the IP 10.0.0.5 that appears in the ipaddrshow is wrong, the IP of the internal interface to the remote CP depends on the swith model, being 127.1.1.5 in this case. But I consider that It is better to have any lan problem fixed first, instead of forcing a failover.

Rgds,

Felipon

dmj
New Contributor
dmj
Posts: 4
Registered: ‎11-01-2012

Re: DCX failover

Thank you. I was thinking it's intended behavior "by design", just wasn't sure.

New Member
ClayCombs
Posts: 1
Registered: ‎06-19-2008

Re: DCX failover

I am concerned with this too.  We are now moving to a remote datacenter and I definitely would like to have the ability to logon to the standby CP and issue hafailover to gain access again.

Join the Community

Get quick and easy access to valuable resource designed to help you manage your Brocade Network.