nt38 11/28/05

[supportsolution]

Hello,

I only pointed out what I read on the forum, forgive me if this is wrong.

Yes I am a paid customer, and about to move a dozen or so websites to your hosting. I could post "hello" in each forum to up my stats if you like ?

To move 10,000 or so files between servers I use software like CA arcserve backup. All you need are "agents" on your servers. (linux, unix and windows are supported. MSSQL, MySQL and other agents are supported)

Also, the cost of your hosting is fantastic, that's why ppl move to you, but if your $35 give you less than you 99.9% uptime you quote then you can see why ppl get upset.

The reference was about high availability, not the technology involved.

Again, don't get me wrong, you do a grand job, but on this occasion it looked a bit shabby from a customers side.

Regards

Dave.

[kattouf]

:-( i was really hoping the new backup system would be better than this. It is just taking too long.

[Vantage]

We are on T at the moment.

As for faster backups.. Recomend a way to move the data. We have tried every way to move data out there. There doesnt appear to be any way to move data onto a Windows server any faster than the way we are doing it.

[powvex]

Raid 5

[serverzb]

wow, this has not been resolved?

Raid 5 might be the way to go as it will cost LESS than this disaster will cost you

[Vantage]

Well, we lost 2 hard drives. WIth raid5 we would need at least a 5 drive array to recover from this.. Putting a 5 drive array in each server plus the raid5 capable card would not be way to much money.
And the general consensus here is that (as 2 drives were lost) it was either the raid card or some other hardware issue that caused the 2 drives to die. SO... raid5 wouldnt have helped at all.

Plus the write speed issue from raid5 would mean that restoring the data would take even longer.

[Vantage]

On V at the moment.

[craigrwarner]

Hopefully the process will be improved to include setting file permissions during the restore. Now that the restore is complete for my sites, file permissions are wrong making the sites inaccessible. So now my sites are down until all restores are done and a script is run. Ugh.

[kapuwa]

i see webalizer stops on 22nd - is this a backup 7 days old?

[Vantage]

You should be current up to the 24th. If you have a webalizer issue, submit a ticket. there are bound to be a few issues with starts programs.

[Dean]

Copy of user data is done, 90% of sites should be online and working with out issue now. We are now resetting permissions for each users back to H-Sphere default permissions. Once this is done if you need any custom permissions we can reset them but not until the script as done as it will just change our settings.

[kattouf]

One question: Do you have a separate network for backups or do it over the same one?

Because i read somewhere, that large networks have 2 network cards in each server and do backups on a dedicated network. Would that be faster.... or?

[byron]

The initial file and directory restoration from backup process has completed. We are now running scripts to restore proper ownership, NTFS permissions and ASPNET settings to all directories. This could take some additional time.

[CreativeNRG]

Is the ownership problem the reason why we can't FTP to any of our directories at the moment ?

Thanks for your help on this - I only hope we are very nearly the end of the disaster so that I can go back to my clients with something positive to say and find out how much this is actually going to cost me because of the downtime.

[Brangwyn]

Originally Posted by powvex

Raid 5

Probably a bad choice for a webserver, Raid 5 has fairly poor performance characteristics (theres always a trade off, more redundancy generally less write efficiency) can be improved from stripe tweaking but thats a PITA as well. Raid 0+1 would be a better choice but that still requires 5 Drives per machine.

My product of the month for backups that I use for all my SBS installations is EMC Dantz Retrospect http://www.dantz.com/en/products/index.dtml Set and forget, backs up *NIX as well and I suspect I've recommended it in the past too. There is a eval version available if anyone feels like checking it out. Replistor is quite a nice product of theirs too.

On my own servers, generally I have just a second drive NOT in a mirrored configuration, but on a controler independant of the primary drive and I simply use a Daily disk mirror job to this drive, then Retrospect off this drive my daily backups, means two levels of redundancy and unless it's a corrupt controller the backup drive usually is always available, you can also spin-down that secondary drive when not in use to preserve it's MTBF.

Because i read somewhere, that large networks have 2 network cards in each server and do backups on a dedicated network. Would that be faster.... or?

It's generally good practice to have one NIC per CPU in a machine for balancing purposes, having an internal network connection can make backups a lot more efficient, and things like traffic to from internal DB servers etc will benefit too.

Is the ownership problem the reason why we can't FTP to any of our directories at the moment ?

Yes permissions needs to be done before things get back to normal.

You should be current up to the 24th

I take it thats confirmation that the previous 3-4 days backups were indeed corrupt/not available? .. not good.

With the revenue you loose each time this happens, hiring a certified disaster recovery consultant to come in and take a look at things would probably pay for itself pretty quickly. Someone who knows all the products out there, their capabilities and shortfalls and who can suggest and implement a failsafe BRP strategy for you.