//*
Win an LCD monitor at the Blades Buyer's Guide Wiki Project
Learn how:
*//
Message
>Hello all,
>
>I'm doing testing HACMP failover for shared volume groups by
disconnecting the fibre cables to the node with owning the
resource groups. The backend is EMC CX and Powerpath is managing
the multipathing (I'm not an EMC person).
>
>IBM support says that this is correct (waiting on the email
explanation)due to the multipathing driver not releasing the
path. There may be a way to script this and I'm hoping that
someone bright person out there will already have something in
place that I might be able to implement. Any help would be
appreciated.
>
>thanks,
>
>james
James, you haven't fully described your environment, so I'll have
to make some general remarks. First of all, you must recognize
that you have already invested in fully redundant components for
storage and the SAN, so there should be no dire consequences to
losing a single storage adapter or cable (i.e you should only
temporarily lose one path) on your HACMP node. Second, you
should recognize that HACMP does not monitor VGs can
monitor applications (which typically don't fare well without
their VGs) if customized to do so. The only thing you get TB
with HACMP is an AIX Error Notification Method that will perform
selected fallover when a shared VG is lost, but only under
certain circumstances (e.g loss of a mirror copy). If your
configuration is like the one I worked on last year (EMC RAID 5
LUNs without AIX LVM mirroring), you'll not see the AIX error
that triggers that AIX ENM, even when pulling both FC cables at
the same time. It is possible to monitor (custom!) the EMC paths
employing EMC utilities and react accordingly, but you run the
risk of false failovers if your monitor method goes haywire. And
of course, you'll have to maintain the custom monitor that you
write. And don't lose perspective that HACMP's design point is
typically to protect against single points of failure; this
extreme scenario is a dual simultaneous failure within your SAN.
My advice to you is to trust the investment you have already
made in the SAN/storage infrastructure to achieve high
availability in your storage architecture through component
redundancy and multi-path device drivers. Confine your HACMP
responsibilities to monitoring those things which HACMP does
well (network failures, node failures, hung nodes, application
monitoring, etc.).
//* INTERESTED IN THIS MESSAGE? RELATED CNTENT
Access Archived Data in the Blink of An Eye (White Papers)
Lotus notes / Domino easily integrated as a BI datasource (Blogs)
IBM AIX (Wiki)
Re: VPN Solution for Redhat Linux (Groups)
*//
//* RESURCES FR THIS GRUP
View :
Manage Account/Group Tips: http://ITtoolbox.com/da/228578
Wiki:
Leave This Group: mailto:leave-ibm-aix-l (AT) Groups (DOT) ITtoolbox.com
Copyright (c) ITtoolbox and message author. No redistribution.
*//