multistore how to create a vfiler with existing server still exists

imarcroft · ‎2010-09-22

Hi I am currently setting up multistore.

The customer has existing windows file servers and wants the vfilers to replace the existing windows servers, so they don't need to modify login scripts to shares and manual mappings.

I was planning to create the vfiles and associated shares, migrate the data then change the vfiler name to reflect the old windows file server name once they are decommisioned.

Has anyone played with this before, and is my plan above a sensible one.

Many Thanks...in advance

scottgelb · ‎2010-09-22

Absolutely! This works very well... you can also just use netbios aliases to respond as each server without vFilers, but that doesn't take care of duplicate share names between servers (which you would have to change if duplicates without vFIlers or implement dfs). Using vFilers, you can create a vFiler that will respond to the name of the server and contain all of its shares without any changes to the server share names. You can even use vfiler add -i and add the IP address of the replaced server if you want to re-use the IP of that server.

Take some time to look at the networking though...to make sure you have enough bandwidth on the FAS to handle all the servers...and if you need separate IPSpaces or not. Often in these brute force use cases of MultiStore we find we can share an IPSpace since all vFilers replacing servers are on the same network. If a shared IPspace, there can only be one default route for that namespace, so if different networks are needed per vFiler look if you need net/host routes for a vfiler. Any vfiler routes need to be in the physical controller vfiler0's /etc/rc file since a vfiler doesn't have an rc file and you use "vfiler run vfilername route add" in rc to add the route for those vfilers.

imarcroft · ‎2010-09-22

Hi

Thanks for the response.

I am using the snapmirror replicated configuration, this automatically replicates the associated volumes attached to the vfiler.

there is a lot of messing around tho, creating, then renaming the vfiler, then having to rerun cifs setup when the vfiler name is changed.

I wondered if it may be better to create the share off the main filer until the data is migrated, then create the vfiler after and associate it to the exiting volume share containing the migrated data?

scottgelb · ‎2010-09-22

Unfortunately, if the volume is owned by vFiler0 and there are shares to those volumes, you have to delete ALL the shares to those volumes before you can add (move) the volumes to a vFiler. So, I would create the vFilers up front if you want to have the shares pre-created. You could break the mirrors, create the vFilers, then reestablish the mirrors (to the vfiler or even vfiler0 still works) then create the cifs shares in the vfiler.

imarcroft · ‎2010-09-30

Thanks for the response to this.

I am going to create a vfiler with a different name to the file server I am migrating, configure CIFS. I won't run the vfiler dr configure command at this point, instead migrate all data.

then when the file server is decommissioned, run vfiler rename oldname newname. then rerun cifs setup to configure the netbios name (after running a cifs terminate)and add the vfiler to the domain.

I will then run the vfiler dr configure command, to setup the snapmirror and failover.

I will then change the ip address of the vfiler

vfiler add vf01 -i 192.168.x.x

and remove the old temp ip

vfiler remove vf01 -i 192.168.x.x

obviously the /etc/rc file will need cleaning up after

thanks

scottgelb · ‎2010-09-30

That works..and some other things you can consider.. vfiler dr configure and vfiler dr resync set the snapmirror.conf schedule to 0-59/3 to update every 3 minutes... no one keeps this default so update snapmirror.conf after. If entries already exist for those volumes, save the file off (before any dr resync for example). Also, if you want to create the mirrors before the cutover, you can do that then don't run vfiler dr configure. The volumes can belong to vfiler0 on the target until you are ready to setup dr. The workaround is to break the mirrors on vfiler0, manually create the vfiler on the target ("vfiler create -r vfilerrootvol" and only have to specify the root volume and it will find all other volumes for the vfiler), stop the vfiler, the do a vfiler dr resync. Then you have the mirrors already over way before the cutover and can have a dr vfiler one snapmirror incremental update away from the source.

ericbottmtm · ‎2011-02-28

I'm preparing to go through the same process you have outlined here. I know the post is several months old, but I'm wondering how your migration went and if you ran into any problems along the way.

Thanks for your time.

imarcroft · ‎2011-03-01

HI

you can change the name and IP address of a vFiler, it went ok.

filer1> vfiler rename vf01_NEW vf01 (vf01_NEW was its temp name)

filer1> vfiler run vf01 cifs setup

then change the IP

The new IP address must be added before the old one can be removed.

filer1> vfiler add vf01 -i 192.168.116.141

filer1> ifconfig ns0 -alias 192.168.116.131

filer1> vfiler remove vf01 -i 192.168.116.131

clean up the \etc\rc file

you need to run through the setup to bind the IP address

filer1> vfiler run vf01 setup

you now need to delete the DR Vfiler, as you can't change the IP address in this state.

filer2> vfiler dr delete vf01_NEW@filer1

this will break off the snapmirror (may want to take a copy of snapmirror.conf)

A new vFiler with the correct name should be created, each volume in the vFiler will need to be added and not just the root volume as below

filer2> vfiler create vf01 -r /vol/vf01_root

you need to bind the IP address again to the DR vfiler, to do this you will need to stop the primary vFiler

filer1> vfiler stop vf01

filer2> vfiler run vf01 setup (to bind th IP address at DR)

now stop the DR vFiler

filer2> vfiler stop vf01

now resync the DR Vfiler

filer2> vfiler dr resync -l root:netapp123 vf01@filer1

a vfiler status command on each filer head should show the correct IP address bound and the DR vFiler is in the correct status

Hope this helps

scottgelb · ‎2011-03-01

All good..great seeing more and more vfilers :). Here are some workarounds. You can ifconfig the .141 address after adding to the vfiler without setp. It will bind itself to the vfiler since it matches the unconfigured ip. Setup works (setup -e) but wacks several files (hosts, hosts.equiv, exports, resolv.conf) so I prefer just ifconfig and skip setup and having to fix files from bak versions.

The dr vfiler could be activated, change the ip (add then remove exactly as you described), then ifconfig then dr resync... Watch out to re-edit snapmirror.conf after configure or resync since it updates to 0-59/3 for every 3 minutes.

Typos Sent on Blackberry Wireless

ROBERT_ACI · ‎2014-01-06

Scott, I sent you a message on this. I've created a similar environment however I'm unable to map to the FQDN \\scantest. I can however map to it's IP \\167.171.83.60

Any suggestions?

Thanks

scottgelb · ‎2014-01-06

Name resolution...can you ping scantest and check dns or host file settings then flushdns on the host also.

ROBERT_ACI · ‎2014-01-06

C:\>ping scantest

Pinging SCANTEST.palmettohealth.org [167.171.83.60] with 32 bytes of data&colon;

Reply from 167.171.83.60: bytes=32 time<1ms TTL=254

Ping statistics for 167.171.83.60:

Packets: Sent = 3, Received = 3, Lost = 0 (0% loss),

Approximate round trip times in milli-seconds:

Minimum = 0ms, Maximum = 0ms, Average = 0ms

C:\>nslookup scantest

Server: dnscolumbia.palmettohealth.org

Address: 167.171.128.1

Name: scantest.palmettohealth.org

Address: 167.171.83.60