Re: majority node set cluster questions

Tech-Archive recommends: Repair Windows Errors & Optimize Windows Performance

The risk that that hardware fails or that the file
system becomes corrupted is too great -- we've had both
happen

What caused the file system to become corrupted? What
hardware were you using in that case?


100M files on a single volume with Win2K. It's fairly
consistent -- when you approach that number, things start
getting unstable under load, writes start failing. You'll
have to take my word on this because I don't believe
you'll find anyone anywhere who has run 100M files on one
volume under 2000 besides us. MS told me they had never
tested beyond ~20M files on one volume under Win2K. I've
since been told that under Win2K3, they test with 20M
weekly and with 100M at milestones. But I don't know the
details, how write-intensive the tests are... this is a
write-intensive environment.


Right now we have our app write each file to two
separate servers, \\fileserverA and \\fileserverB,
considered by our app as primary and standby.

I take it that you haven't heard of backups. <G> How
about back ups to disk and then to tape? How about DFS?
What data is your app writing? Maybe it would be better
if it were writing to a SQL database.


it would be better going to a database. Most of it should
be going to one. We are rewriting it, but it takes time.
It's very complex.

so what we need to do is rid ourselves of the concept
of primary and secondary storage but still have two
separate and distinct storage repositories on the back
end. The cluster would simply write to both
simultaneously instead of replicating from one to the
other.

This is not going to happen with Microsoft products
only. You can look into 3rd party stuff, but it sounds
to me like you have a poorly architected application and
have not really looked at this as a DR concern.

yes, it is poorly architected. But we didn't build, we
just have to deal with it.


Sorry, this is not how MNS works. You also need to
remember that MNS is for the quorum only and does not
provide similar support for data disks.

I didn't think this was possible with MS clustering and
only recently found it might be a solution.

also, is there any limit to the number of shared
resources i could have in this scenario given my use of
UNC paths?

I think you should back up to square one and start with
what the business requirements are for the application
and whether the application is meeting those business
requirements.

unfortunately, not an option. So i have to come up with
our -best- option that gets us to the point where the app
has been rewritten. It may not be ideal, but i've got to
have two copies of that data since the possibility of file
system corruption exists as long as we are pushing it.. i
have no problem whatsoever with replication like
DoubleTake or even DFS -once the app does things more
appropriately-. For now, they're risky and we're better
off having the app write separately, if that's the only
option.

thanks for your replies.









.

Relevant Pages

  • Re: How to corrupt a MS Access database file in a normal way?
    ... Log when they start the app, log when they stop the app. ... My problem is how to generate the corruption by common termination of a ... but bad versions of Jet on the machines using the VB app ... the corruption problems could be caused by the software ...
    (comp.databases.ms-access)
  • Re: Corruption - how to MAKE it happen?
    ... It's just that I *really* have this feeling that the corruption is ... Obviously if you've designed your app so that it takes so long to ... You've done something to speed up the load time!" ... and it made them feel like the app was loading more quickly. ...
    (comp.databases.ms-access)
  • Re: Does Access Break?
    ... Larry, you're using a completely different definition of "break an ... so I consider corruption issues to be failures *outside* ... itself -- it's a breakdown of the operating environment. ... If the code in an app works and remains unaltered, ...
    (comp.databases.ms-access)
  • Re: vfs::zip corrupts memory?
    ... Reading from zip files on other parts of the app works. ... As the proverbial memory corruption problem, ... executing the tcl code from tclsh ...
    (comp.lang.tcl)
  • Different same machine browser sessions intefering with each other with Forms Authentication and asp
    ... separate IE processes and starting two windows in the same IE process ... We have an asp.net 2.0 app that uses formsauthentication. ... I think there are other symptoms of one session interfering ... tried messing with the local IE cookie settings and it didn't make a ...
    (microsoft.public.dotnet.framework.aspnet)