nt38 11/28/05

[Shep2113]

Any more news on this yet?

And one other possible idea, zip without compression, just to make one file to transfer for backup? Less cpu time needed that way, maybe that would help?

-Eric

Doh! I was behind you by a minute

[Brangwyn]

When we replaced the drives in NT1 NT2 and NT3 the other night I think it was we copied from old drive to new drive this still took 8 to 10 hours.

Was the server serving pages at the same time? just curious ... earlier I was talking about it being probably a little faster if the server wasn't running, less disk trashing should occur imo .. Dean or someone mentioned it was good to have the server giving a default page to let people know it's at least up, which probably is a good idea .. just maybe combining the two ideas, you could have a "standby" server that simply just dishes up a message saying server down for maintenance and route all requests for the server being rebuilt to this standby one whilst you do the maintenance. ... talking out loud here I guess.

Still 8-10 hours surprises me, I rebuilt a 120GB Server on the weekend, took about 5 hours total and that included doing a rebuild of the OS, then Retrospecting back about 75GB of content over the network from my storage array.

[kapuwa]

Brangwyn - Please help them, they need advice from people like you now. This is taking way too much time and can no longer bear the pain.

[Shep2113]

Yeah that was my thought a few months back. Just a server that can step in and say 'this site is experiencing problems, please try back later'.

They explained to me it's not that simple though, because of routing and other things. I don't remember all the details, but it sounds easy from the outside but making it happen is a much more difficult task.

It'd be nicer than seeing NT38 up in the corner though, if it were possible.

[byron]

12:30 Update time! Sorry, I'm a little late.. phone calls and tickets are heavy right now, which is understandable considering the situation at hand.

Okay, so far the recreation scripts have completed setting ownership permissions on files and folders, but another script responsible for NTFS file permissions is still running. We are also still watching the script that recreates all ASPNET permissions and functionality across sites, as it is also still running. If you are calling regarding NT38, this is all I'm going to advise you of over the phone--unless we are working together on something specific within your site that needs attention.

If you are trying to call the helpdesk now and you're just getting a ringing phone or voicemail, you are more than welcome to hang up and try calling again but *please* bare in mind:

Many NT38 resolutions are not going to be as simple as making a phone call, so please try to open a support ticket with us instead. There are other callers with unrelated issues that deserve just as much attention as you do.

[byron]

Okay, all scripts are done as far as permissions recreation are concerned. I'm now going to start working on ASPNET/Access DB trouble tickets that have been stacking up. I will also be restarting the server as necessary between 1:30a - 2:30a to resolve any outstanding issues with file copies or other things that keep us from getting permissions working properly. I will post further information when I have some.

[byron]

I've gotten a couple calls on this one already, so I wanted to post. If you've previously had us create an "Application" status for a site directory in the past before this incident on NT38, please send us a ticket requesting the creation once again. These are manual recreations and must be requested again by you.

[byron]

I am applying ASPNET user permissions to user directories now. This was not done during the previous script processes, and is still necessary for most ASP .NET sites to function properly.

[supportsolution]

Hi.

Obviously you have a server farm with different OS's and different tasks.

Here is an example for a good resilient server;
with good IO speeds
with good recovery times
With separate system and data drives
with redundancy on both drives
with managed mirroring. (gives you 24 window to swap)

Database Server.

Server build.
Hardware SATA raid will do for this. (keeps cost down)
Array 0 - System Drive (2 drives 60GB)
Array 1 - data drive (2 drives 250GB)
Raid 1 (Mirrored)

Create your array0 (System) that mirrors to your other drive within your array.
Create your array1 (Data)that mirrors your other drive.

Install you OS. Do a full, not quick format of your System drive.
When OS installed format your datadrive (again, full format not quick)

Install all your apps
disable all services (windows) daemon (unix) that are not required.

Defragment your HD's

* at this point your server is ready to serve your clients.

Backup your server with a suitable backup software. Arcserve v11 is good, you could go with CommVault. AND Use Ghost. Make a ghost image of your newly created server. burn to 2 DVD's. 1 for archive, 1 to be taped to server.

Put server onto domain.(Windows, if required)

Allow the server to serve.

Backup routine set to GFS or incrementally with synthetic fulls on Friday.

The cost above would be 2x hard drive costs.
OS downtime. 1 hour using Ghost or 1 hour using a DR backup agent for Arcserve
Restore time. .5-2 hours depending on backup software and data size.

-=-=-=-=-=-=-=-
Using backup software that support FS (file system backup devices).

Pro
Huge backup / Restore speeds

Against
You need a huge data storage area.
your backbone is 1GB or you use 2 nics. (2 separate subnets)
No off-site storage

-=-=-=-=-=-=-=-
Using tape backups.

Pro
off-site storage
separate media for separate jobs

Against
Large cost for drive and tapes
Either you have a tape monkey or a tape library. (more cost)

-=-=-=-=-=-=-=-
Using Tape backup's with FS

Pro
off-site storage

Against
Large cost for drive and tapes

-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*

This is just 1 way to swing the cat, but there are others

Robocopy is not the way to go. you do not want a file-copy as a backup.

Pro
Free

Against
OS overhead - the server will try and use as much CPU power as possible.
Users must be created on server before a robocopy is "restored" or the ACL's will be corrupt.
A full HD defragmentation must be completed before copy and after a copy is completed.
bottle necks are; CPU, HD controller, HD drive.
Robocopy is not a "managed" product. backup software is.

+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-

Hope this helps.
That should cost ya if you where a client lol

if you use the above, please deposit money into my paypal account

Regards

Dave
Geek ? no - perfessional ? no well..... yeah

[byron]

ASPNET permissions are still applying themselves to user and system directories.

[CreativeNRG]

How long will this take and is this why I am getting web.config errors for no reason at the moment ?

[jjay]

7:28 status report plz?

Rick

[supportsolution]

Hi,

With regards to a default page on a website.

All website "names" are forwarded to an IP address.

An IP address is "hosted" at Vortch's RIPE allocation. (a list of IP addresses that are listed at Vortch offices associated with there router)

You (your customer etc) types in www.fathairywoman.com and that domain name gets turned into an IP address (81.81.81.81 example) and sent with the domain name header to Vortch. The IP address is received and passed to the server hosting the ip address (81.81.81.81). The IIS server looks at the header (www.fathairywoman.com) and passes to the right website within IIS.

Now.....
A server that has gone down. (a)
81.81.81.81 hosting www.fathairywoman.com! you get a 404 error or a msn search page. etc... bad for us (resellers)

To do.... (vortch staff)
A server that is not doing much (b) (either clean or not hosting a lot) gets given a second (or third) IP address within its bindings. (very easy to do) 81.81.81.81

You create a new website within IIS and point to the documents folder that contains default.htm explaining the website is under maintenance. (unbranded)
You edit the IIS website and change the header information to be the IP address only. (this means it don't matter what the host header was. (i.e. if www.fathairywoman.com, www.google.com, www.Ilovesheep.org are pointed to the same IP address it doesnt matter)

The all requests for 81.81.81.81 will now get the maintenance page

Then when the downed server is up and working again (a) (under a different IP address so it don't conflict)
You remove the IP address of the server that's not doing much (b) 81.81.81.81
You change the IP address of server (a) back to 81.81.81.81

Jobs a jolly one!

No routing needs to be changed.
No DNS needs to be changed
Creating of website on webserver 1 min max!
Happy resellers and clients - priceless

Again, post money to my paypal account.

Regards

Dave

[CreativeNRG]

What on earth is going on with NT38 now ????
It is showing critical errors again and sites are not accessible, this is now getting beyond a joke !!!!

[pentom]

FYI...
Anyone using SSL probably needs to reinstall their certs...I had to.