UNSOLVED

HabibG2

updated

19 years ago

H

HabibG2

455 Posts

0

1252

February 4th, 2008 06:00

Odd behaviour when backing up Physical Nodes.

I have a simple file cluster platform with 2 Pnodes and one Vnode,

I notice when I perform a backup on one Pnode, when backup starts on the local SN to that Pnode, then halfway through it moves to the SN on the second Pnode. The Storage Node field on each Pnode had the local SN and "nsrserverhost". What could be the causing this?

Also I notice when I backup one of my Pnodes, some sevasets remain pending, but the job terminate with "write Completion"and volume ejected and no errors. If I restart the job then the backup terminates ok. I am sure there is a reasonable explanation for this behavior, but I can seem to find it.

Thanks for your help.

Habib
  • amediratta

    2 Intern

    •

    2014 Posts

    938

    0

    Posted February 4th, 2008 07:00

    Are your savesets for the physical node only. Nothing from the shared resource. Also keep only the same physical node as the storage node attribute.
  • HabibG2

    455 Posts

    938

    0

    Posted February 4th, 2008 09:00

    Yes. On my Physical nodes I am only backing up;
    C:\
    Asr
    System DB
    ....etc

    I do keep the same physical node as the Storage Node, as I explained ion my question.

    I backup my shares on the Vnode.

    Regards
    Habib
  • ble1

    6 Operator

    •

    14354 Posts

    •

    56186 Points

    938

    0

    Posted February 4th, 2008 14:00

    Does it happen after 30 minutes?
  • HabibG2

    455 Posts

    938

    0

    Posted February 4th, 2008 23:00

    No. My backup does not even last for 30 mins. Its is also not consistent, as change SN can happened at various stage of the backup and for different part of the savesets. I have used PATHIGNORE on both by Pnodes.

    Regards
    Habib
  • ble1

    6 Operator

    •

    14354 Posts

    •

    56186 Points

    938

    0

    Posted February 6th, 2008 14:00

    Let?s talk about your setup... I'm a bit confused with it... you said your physical nodes have nsrserverhost value, but they are also storage nodes - is that right? Also, why do you use pathownerignore? (not that I believe it is related, but I just wonder). I'm trying to figure out what is going on, but I have problems understanding your setup... Can you explain a little but more?
  • HabibG2

    455 Posts

    938

    0

    Posted February 8th, 2008 03:00

    Apologies for the confusion!!

    Basically I have 2 physical nodes FILE001 and FILE002 and one Virtual Node FILECL

    In the Storage Node field of FILE001 I have;
    FILE001.domain.com
    nsrserverhost

    In the Storage Node field of FILE002 I have;
    FILE002.domain.com
    nsrserverhost

    In FILECL the storage entry are;

    FILE001.domain.com
    FILE002.domain.com
    Curphyhost

    Both FILE001 and FILE002 are SNs not DSNs. I include "nsrserverhost" just in case there is a problem with the local SN itself ( this is what I was told by an EMC guy).

    Is this any clearer?

    What I don't understand, why the backup move between the SN of FILE001 and FILE002.

    Regards
    Habib
  • HabibG2

    455 Posts

    938

    0

    Posted February 11th, 2008 01:00

    I will check , make amend if its not, and will confirm.

    Ta.
    Habib
  • ble1

    6 Operator

    •

    14354 Posts

    •

    56186 Points

    938

    1

    Posted February 11th, 2008 01:00

    FILECL should have curphyhost listed first.
  • HabibG2

    455 Posts

    938

    0

    Posted February 11th, 2008 05:00

    Forgto to mention.

    The behoviour is being experienced on the Physical Node FILE002, not the virtual node. I placed Curphyhost first on the virtual; node.

    Rgds
    Habib
  • ble1

    6 Operator

    •

    14354 Posts

    •

    56186 Points

    44

    0

    Posted February 11th, 2008 06:00

    Ah ok... curphyhost should be used only with virtual node.

    As for file002, I really don't see why would that host use file001 as it is not mentioned in storage node list... unless there is some mess in the background with names/aliases and addresses. That would be the only explanation.