Trying to find an optimal solution for a client with various problems!
-
One of my clients has an office with around 25 users. 3 years back, I was asked to setup a server with central file sharing and permissions and had limited budget. So I setup one Windows server with AD and file sharing. Data being backed up to crash plan and works well, several situations CP helped them to recover deleted files Works fine till this time, but few things that they want to improve and sort few issues.
Few weeks back, they called me saying no one can access the network drive/server. Upon checking found that the server's on board network card got fried! Replaced with an additional card and things started working again.
Before I go further, let me give you some details about their office.
The office is in a warehouse, almost in the middle of desert, so very dusty, hot and during winter/rain water can come in to the office, so kept the server in the first floor to avoid any damage due to water. Other issues they have are sudden power spike & failures, had blown out their fridge, washing machines and PCs. (I had an APC tower UPS for the server with spike guard so the server was protected till now, seems like that's not functioning very well now) On top of this, their internet connection cable comes through a man hole which sometimes gets filled with water, damaging the cable, and ISP comes and fixes it once in a while. No option to get alternate ISP there, tried USB dongle wifi, doesnt cater to their usage.
Sounds like a perfect setup NOT to have an office there but they currently don't have any choice! May be they will be able to move out to a new office next year, but for now this is it.
Now, they've asked me to setup a fail proof solution to make sure that even if one server stops working, they want to continue to work. The obvious answer was to setup an additional DC and move this data to a network drive (DROBO!). One thing they didn't wanted to spend a lot on storage now, as its almost end of the year, but they are not completely ignoring it.
Even with a network drive, they fear of that device getting damaged, with DROBO, i have the drive protection but wondering if they have any enclosure warranty/ may be with an additional charge, get an enclosure on call to be replaced in case this goes bad.
Other option i am thinking of is to have the second server, add internal hdds on that and may be do a sync between the server1 and server 2 for data. So if one goes down the other can take over, and data can be served from there.
I also suggested them a hybrid option, may be have the data synced to an online solution, so the users can switch to that if the server is not working, but again since the internet is not very reliable, this might not be the best solution.
What are the best options to give them a cost effective and fail proof solutions in this case?
-
You're definitely between a rock and a hard place with this one.
They tell you the goal is to continue working in case of a server failure. Sounds to me like only a second server being synced with the first will give then what they want.
DFS should do what you want assuming you stay with Windows. Rsync (I think) could work between two linux servers.
A Drobo would be a single point of failure. This would not give them what they asked for.. if the Drobo dies, then the whole thing is down.
Does the client have any type of wireless access to internet options available to them?
-
@Dashrender said:
A Drobo would be a single point of failure. This would not give them what they asked for.. if the Drobo dies, then the whole thing is down.
And is far more likely to fail than a server AND takes longer to fix (you can't just go swap in commodity parts.)
-
I am thinking about using DFS/ a sync tool, but they also have quickbooks and another software whose db file is saved and needs to be accessed from the network. Need to test the db access via the software using a share name rather than the drive physical path. Does the file sync via DFS or other tool have any impact on the network, for users accessing the files?
-
@Ambarishrh said:
Now, they've asked me to setup a fail proof solution to make sure that even if one server stops working, they want to continue to work. The obvious answer was to setup an additional DC and move this data to a network drive (DROBO!).
How is that an obvious answer? Isn't it the opposite? It would take all the problems that you have today and just make them worse without solving any?
-
@Ambarishrh said:
Even with a network drive, they fear of that device getting damaged, with DROBO, i have the drive protection but wondering if they have any enclosure warranty/ may be with an additional charge, get an enclosure on call to be replaced in case this goes bad.
You have drive protection in even the most entry level server. Drobo does have drive protection, but very little. It offers only RAID 5 and RAID 6, that's all. No RAID 10 option, for example.
Having a second Drobo empty and standing by is an option not a very good one. You could use a normal server and replace parts easier and more cheaply. Drobo makes nice gear but is the exact opposite of what you want there. When "reliability is the key concern, Drobo is the last device on your list. Drobo is for tons of low performance storage at very low cost where reliability isn't important. It's great for backups, archives, etc. Never for production storage. And you are dealing with a situation where normal production storage isn't enough and you want to improve things, so you are going even farther from a Drobo use case.
-
@Ambarishrh said:
Other option i am thinking of is to have the second server, add internal hdds on that and may be do a sync between the server1 and server 2 for data. So if one goes down the other can take over, and data can be served from there.
This is actually the only option. External storage is never an option until you have four or more servers and never reasonably one until you have about ten. At one to two, never would external storage be the slightest option even when reliability doesn't matter.
-
How rapidly does the data change? If the data is not that high of priority (you can lose a few minutes, you can have a few minutes of downtime) then Linux with RSYNC is ideal. Or you could use ReadyNAS or Synology with RSYNC too, same thing.
If you have high priority data and it changes very quickly, then Linux with DRBD would be the answer.
-
@scottalanmiller said:
@Ambarishrh said:
Now, they've asked me to setup a fail proof solution to make sure that even if one server stops working, they want to continue to work. The obvious answer was to setup an additional DC and move this data to a network drive (DROBO!).
How is that an obvious answer? Isn't it the opposite? It would take all the problems that you have today and just make them worse without solving any?
How does adding a second DC adds more problems? The issue the client have now is the single point of failure. Adding a second server with the file sync solved this issue rite?
-
@Ambarishrh said:
How does adding a second DC adds more problems? The issue the client have now is the single point of failure. Adding a second server with the file sync solved this issue rite?
Because you are adding an inverted pyramid of doom with the point of the pyramid being even MORE fragile than what you have now. So MUCH less reliable not just a little less.
You wouldn't reduce the single points of failure, you would be making the one point of failure less reliable and adding additional points of failure. More failure, zero protection.
It's not the second DC that causes the issue, it is how it is added (Drobo SAN.) DCs should never, even when they are in massive environment, have external storage. When you do that you are undermining their built in reliability. Even if you have SANs, you make sure that your DCs don't talk to them.
DCs need NOTHING to make them highly reliable. You never sync a DC, you never put it on external storage. You just have two and keep them on local storage. DCs automatically make themselves highly available with no additional interaction.
-
@Ambarishrh said:
Adding a second server with the file sync solved this issue rite?
Nothing involving file sync should be involved when talking about DCs.
-
@scottalanmiller Seems like there was a confusion. I had 2 scenarios;
1 adding a second DC and keeping the files on the internal drives of both servers, sync between the servers.
- Adding a Drobo. this was considering the fact that Drobo gives additional protection for the storage.
In this case, i am thinking of having an additional DC with an internal storage and DFS enabled so all gets synced between the two servers. Does it sounds ok ?
-
@Ambarishrh said:
1 adding a second DC and keeping the files on the internal drives of both servers, sync between the servers.
This should never happen. Never. Two DCs, let them take care of themselves.
-
@Ambarishrh said:
- Adding a Drobo. this was considering the fact that Drobo gives additional protection for the storage.
Adding a Drobo would take away protection, not add it. I'm not sure what you are imaging a Drobo is, but it definitely does not do what you are picturing. It's the opposite.
-
@Ambarishrh said:
In this case, i am thinking of having an additional DC with an internal storage and DFS enabled so all gets synced between the two servers. Does it sounds ok ?
DFS is for file serving, DC is for Active Directory. The two should not overlap.
I don't think that DFS is an option with things like QuickBooks.
-
I am talking about the hdd protection on Drobo. Was thinking if i have the data on Drobo, single hard drive failures can be prevented.
-
@Ambarishrh said:
I am talking about the hdd protection on Drobo. Was thinking if i have the data on Drobo, single hard drive failures can be prevented.
Yes, it has very basic RAID. Are you suggesting you are running servers with no RAID at all?
-
So keeping the current AD, what could be the optimal solution that you recommend?
-
@Ambarishrh said:
I am talking about the hdd protection....
This is weird terminology. When talking about this, be sure to call it RAID generically or RAID 6 specifically. Just calling it "HDD Protection" makes it sound like something unique.
-
@Ambarishrh said:
So keeping the current AD, what could be the optimal solution that you recommend?
Add a second server. There is one and only one model for AD DCs. You always run them with local storage, you just add more of them. You never "do" anything to make them reliable. Anything you do will just break the reliability that is built in. You never let them talk to external storage, never let them sync, never do anything special. They are HA natively, just have two of them on completely different servers.