|
Home > Archive > Data Storage > March 2007 > vif takeover scenerio during node takeover
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
vif takeover scenerio during node takeover
|
|
| Raju Mahala 2007-03-09, 1:14 pm |
| There are two filers in cluster mode with below network conf.
Filer1
/etc/rc -
hostname filer1
vif create multi vif1 -b ip e0a e0b
ifconfig vif1 `hostname`-vif1 mediatype auto netmask
255.255.255.0 partner vif1
/etc/hosts -
10.199.41.51 filer1 filer1-vif1
Filer2
/etc/rc -
hostname filer2
vif create multi vif1 -b ip e0a e0b
ifconfig vif1 'hostname'-vif1 mediatype auto netmask
255.255.255.0 partner vif1
/etc/hosts -
10.199.41.61 filer2 filer2-vif1
Now I made changes at filer1 as follwos and rebooted the filer1
Filer1
/etc/rc -
hostname filer1
vif create multi vif-pri -b ip e0a e0b
vif create multi vif-sec -b ip e0c
vif create single vif-main vif-pri vif-sec
ifconfig vif-main 'hostname'-vif-main mediatype auto netmask
255.255.255.0 partner vif1
/etc/hosts -
10.199.41.51 filer1 filer1-vif-main
Now my questions are :
1) Will filer2 be able to takeover of filer1 as I didn't change the /
etc/rc file of filer2 for "partner vif-main". So anybody can explain
how takeover happens for netork or we can say how partner filer starts
responding to clients on behalf of partner node. Whether it counsults
partner's /etc/rc file.
2) Take a case that there was some typing error during modification
in /etc/rc file and after reboot filer1 couldn't come at network so
now can I takeover of filer1 by filer2 and then do correction in /etc/
rc file for filer1.
3) why the syntax 'hostname'-vif1 in /etc/rc file instead of IP
itself.
Thanks in advance.
- Raju
| |
| Faeandar 2007-03-12, 7:14 pm |
| On 9 Mar 2007 10:37:20 -0800, "Raju Mahala" <rajumahala@gmail.com>
wrote:
>There are two filers in cluster mode with below network conf.
>
>Filer1
>/etc/rc -
> hostname filer1
> vif create multi vif1 -b ip e0a e0b
> ifconfig vif1 `hostname`-vif1 mediatype auto netmask
>255.255.255.0 partner vif1
>/etc/hosts -
> 10.199.41.51 filer1 filer1-vif1
>
>Filer2
>/etc/rc -
> hostname filer2
> vif create multi vif1 -b ip e0a e0b
> ifconfig vif1 'hostname'-vif1 mediatype auto netmask
>255.255.255.0 partner vif1
>/etc/hosts -
> 10.199.41.61 filer2 filer2-vif1
>
>Now I made changes at filer1 as follwos and rebooted the filer1
>
>Filer1
>/etc/rc -
> hostname filer1
> vif create multi vif-pri -b ip e0a e0b
> vif create multi vif-sec -b ip e0c
> vif create single vif-main vif-pri vif-sec
> ifconfig vif-main 'hostname'-vif-main mediatype auto netmask
>255.255.255.0 partner vif1
>/etc/hosts -
> 10.199.41.51 filer1 filer1-vif-main
>
>
>Now my questions are :
>1) Will filer2 be able to takeover of filer1 as I didn't change the /
>etc/rc file of filer2 for "partner vif-main". So anybody can explain
>how takeover happens for netork or we can say how partner filer starts
>responding to clients on behalf of partner node. Whether it counsults
>partner's /etc/rc file.
It's been a while since I've set these up but I don't see any reason
why this won't work. It's certainly not optimal (a multi-mode vif
with one interface?) but it should work. All that really matters for
partner takeover is that the partner vif knows the IP used on the home
filer. All the other vif information is specific to the home filer.
>2) Take a case that there was some typing error during modification
>in /etc/rc file and after reboot filer1 couldn't come at network so
>now can I takeover of filer1 by filer2 and then do correction in /etc/
>rc file for filer1.
I can't say for certain because it may depend on the error. If the
failover/partner vif was not modified and was working, it should be
able to takeover and you can access the root of the home filer. If
the partner vif was modified and does not work afterwards you may be
stuck on console with rdfile/wrfile.
>3) why the syntax 'hostname'-vif1 in /etc/rc file instead of IP
>itself.
Because it's easier to deal with names instead of IP's. There's no
other reason for it.
~F
>
>Thanks in advance.
>
>- Raju
| |
| Raju Mahala 2007-03-13, 1:14 pm |
| >
> It's been a while since I've set these up but I don't see any reason
> why this won't work. It's certainly not optimal (a multi-mode vif
> with one interface?) but it should work. All that really matters for
> partner takeover is that the partner vif knows the IP used on the home
> filer. All the other vif information is specific to the home filer.
Sorry boss I couldn't get. Lets ex. filer1 takes takover of filer2,
and /etc/rc of filer1 has "partner vif1" so how filer1 gets IP of
filer2 for its vif1.
- Raju
| |
| Faeandar 2007-03-13, 7:12 pm |
| On 13 Mar 2007 10:10:35 -0700, "Raju Mahala" <rajumahala@gmail.com>
wrote:
>
>Sorry boss I couldn't get. Lets ex. filer1 takes takover of filer2,
>and /etc/rc of filer1 has "partner vif1" so how filer1 gets IP of
>filer2 for its vif1.
>
> - Raju
I don't know if it's mirror'd in nvram or it polls at intervals to
find changes, but whatever it is filer1 has the information for
filer2's vif. All it is is an IP, they already know each others MAC
address.
So when filer2 fails, filer1 advertises filer2's IP and MAC on it's
interface. The switch takes a few seconds to figure out the port for
filer2's MAC has changed, which is the reason for most of the delay on
failover.
The file handles for all connections on filer2 are mirror'd in nvram
which is why you don't get stale handles on failover.
~F
|
|
|
|
|