Re: Settings comparison for cluster nodes.
- From: "Vincentb" <Vincent_borreux@xxxxxxxx>
- Date: 11 Aug 2005 07:56:11 -0700
Cyberstore,
thanks for your answer and your advice. We already foresee a failover
during the installation of a new clustered application (or after the
deployment of any major change) in order to validate that everything is
ok on both nodes. In the future we plan to force an 'automated'
failover once a month in order to ensure that both nodes are at the
same 'level'. But this is mainly an a posteriori check (a kind of
validation the the process is well established and working fine).
However, in order to secure that SWAP, I am looking from a priori
checks to detect any discrepancies before it really impacts our
business.
I am interested in getting more information about the SysCompare tool
you have mentioned. Could you please tell us where I could find more
information about it. I have googled about this without success.
Thanks in advance,
Best regards,
Vincent
Cyberstorme a écrit :
> To compare 2 nodes in a cluster or otherwise, you can use a tool called
> "SysCompare" from MS. As Ryan has indicated, the bottomline to fixing your
> issue is to implement processes so that changes can be controlled. If your
> code changes are for applications that are "clustered", it is good practise
> to fail the cluster from one node to the other and validate that the groups
> come online and stay up, as part of the post-implemention checkout.
>
> "Vincentb" wrote:
>
> > Hi,
> >
> > First of all, I am not a technician. So my question may look stupid or
> > naive to all of you.
> > I work for a company which has to manage itself, for its own business,
> > about 1000 Windows servers split on two sites (Production site /
> > Contingency site). The main OS is Windows 2000 (and still very few NT4
> > / 2003 server).
> > Most critical applications run in a clustered mode (Active for the P
> > site / Passive for the C site). In case of major problem on the active
> > node, the clustering mechanism switches automatically the application
> > to the C-node.
> >
> > So far, so good ...
> >
> > Once we instal a brand new cluster, everything is fine. Both system and
> > application layers are configured the same way following a predefined
> > checklist (same DLL, registry parameters, patches, ...). However, we
> > have noticed several times that this ideal situation tends to degrade
> > with time.
> > This is the case when, for instance, a watch duty team is called in the
> > middle of the night to solve an issue: they fix it on one node and
> > restart the application on that node, but 'forget' to apply the same
> > correction to the other node of the cluster (which crashes of course
> > for the same reason when it becomes active).
> > It also happen sometimes we have to apply fixes to all servers but that
> > some of them are 'forgotten' ... or the install does not succeed and
> > errors are not trapped ...
> >
> > I am pretty sure you all have faced that kind of issue.
> >
> > I was wondering if there was an easy way to compare 2 nodes of a same
> > cluster in order to produce a report listing all discrepancies between
> > the nodes? Does this seem feasible to you? Does such a tool exist? Have
> > you an idea of the main parameters we should compare if we had to write
> > scripts on our own?
> >
> > Globally speaking, how do you ensure that both nodes of your cluster
> > are always equivalent from an system / application (/ hardware?) point
> > of view?
> >
> > Thanks in advance for your suggestion and help,
> > Best Regards,
> > Vincent
> >
> >
.
- References:
- Settings comparison for cluster nodes.
- From: Vincentb
- RE: Settings comparison for cluster nodes.
- From: Cyberstorme
- Settings comparison for cluster nodes.
- Prev by Date: Re: Problem with cluster
- Next by Date: Dive configuration on new cluster
- Previous by thread: RE: Settings comparison for cluster nodes.
- Next by thread: failover notifications
- Index(es):
Relevant Pages
|
