Posted: 1/27/2017 3:44:55 PM EDT
|
I have two identical Windows 2012 servers in my company of 100 users.
The server's only function is to serve simple files to users. Server "A" has about 750GB of files. I would like to configure Server "B" in case the main server goes down "B" would jump in and keep on going seamlessly (with the users never knowing what happened). Would I need third party software to do this? Thanks. |
|
Quoted:
You want scale-out file server. https://technet.microsoft.com/en-us/library/hh831349(v=ws.11).aspx look into this or just DFS and compare functionality. the general consensus is that with SOFSFA is meant for serving application data as its primary function not files. DFS is intended for highly available file shares and branch office file server replication. |
|
Quoted:
look into this or just DFS and compare functionality. the general consensus is that with SOFSFA is meant for serving application data as its primary function not files. DFS is intended for highly available file shares and branch office file server replication. You're right, DFS is likely better for end-user file shares. SOFS would be better if he were hosting VMs on Hyper-V or SQL DBs. Even if it's Active/Passive, all his end-users would experience is a few seconds of lag. |
|
Quoted:
It says Scale-Out File Server is not recommended with Home Directories. My users all have home directories. According to that link "General Use File Server Cluster" might be the way I should go? Pretty much, but I would honestly just recommend using a NAS/SAN or whatever that has built in fault tolerance. With NetApp and others you can provision CIFS shares directly from the filer. The share would be on a volume that could tolerate multiple disk failures, controller failures and the like. The fact that you have 100 users and it sounds like you're not virtualized makes my brain hurt. |
|
The two servers I have now are DELL NX3230 NAS devices (one offsite) using Windows Storage Server 2012 R2 Standard.
It is plenty fault tolerant, but when I have to patch the main server, which requires a restart, no one can do anything file-wise until the server comes back up which means I have to schedule the downtime. It'd be nice if it was transparent to the user. Heck, I'd be happy with just a complete up-to-date (live) copy of all the current files on the second server. Then I could just rename the second server and be back in business if disaster struck the first server. |
|
Quoted:
The two servers I have now are DELL NX3230 NAS devices (one offsite) using Windows Storage Server 2012 R2 Standard. It is plenty fault tolerant, but when I have to patch the main server, which requires a restart, no one can do anything file-wise until the server comes back up which means I have to schedule the downtime. It'd be nice if it was transparent to the user. Heck, I'd be happy with just a complete up-to-date (live) copy of all the current files on the second server. Then I could just rename the second server and be back in business if disaster struck the first server. ick... I guess that's a NAS. Sorry about that. If one of them is remote use DFS namespace with DFS-R That gives you one point of entry and the data is replicated on each. the drawback is that (could be wrong things have been changing recently) but file locking is practically nonexistent with DFS-R but it's more or less a non-issue if you it's just for information worker file access and your namespace is setup appropriately with your shares and drive maps. |
|
Quoted:
ick... I guess that's a NAS. Sorry about that. If one of them is remote use DFS namespace with DFS-R That gives you one point of entry and the data is replicated on each. the drawback is that (could be wrong things have been changing recently) but file locking is practically nonexistent with DFS-R but it's more or less a non-issue if you it's just for information worker file access and your namespace is setup appropriately with your shares and drive maps. DFS-R without the file locking is fine for Folder Redirection of user profiles, as file lock doesn't matter on a users own files. It's when you have multiple users trying to make changes to one file where shit gets annoying and the file lock is necessary. If there are multi-user accessed shares, you can still use DFS-R but you have to go buy yourself a PeerLock or something similar that provides for file locking. If he wants straight up fail over clustering with out shared storage, he needs to get a Starwind vSAN license. Then he can seamlessly failover the servers themselves. |
|
A variation of the following (admittedly cynical) SMB scenario:
Someone had an idea of how to do an "offsite backup" wherein once in a blue moon somebody manually (or by job) copies files across the wire until one day the aggregate storage size gets far too large to both keep track of as a catalog and to reliably send across an SMB pipe. Backups become less frequent and often incomplete and there's no central catalog of what is backed up, where, and to what delta frequency. Inevitably there's a disaster and after much panic and doom, someone has to tell the boss that only about half the files are actually on the "other server" and that the last date files were backed up was 9 months ago. It can very from not nearly that bad to "the other WSS has been sitting over there for three years and has never been used. Quoted:
So what I want to know is why is this second WSS off site, but there's no replication in place? What's it doing off site? |
|
Office Manager: Well I opened this email attachment that looked like an invoice from Amazon and now all my files have .locky at the end of them
Me: When was the last backup of all your data? Office Manager: Well I think we did one a couple of years ago, can you get all my stuff back? Me:Yeah.....it's gone. <insert continued reiterations that the data is really gone> Office Manager: *sobbing uncontrollably* Entire accounting system; gone Every word document; gone Every excel spreadsheet; gone Every pdf; gone 2 months later that same office manager disappeared on some sort of weird drinking bender. True story. Quoted:
A variation of the following (admittedly cynical) SMB scenario: Someone had an idea of how to do an "offsite backup" wherein once in a blue moon somebody manually (or by job) copies files across the wire until one day the aggregate storage size gets far too large to both keep track of as a catalog and to reliably send across an SMB pipe. Backups become less frequent and often incomplete and there's no central catalog of what is backed up, where, and to what delta frequency. Inevitably there's a disaster and after much panic and doom, someone has to tell the boss that only about half the files are actually on the "other server" and that the last date files were backed up was 9 months ago. It can very from not nearly that bad to "the other WSS has been sitting over there for three years and has never been used. |
You know what, I'm just going to go ahead and post this video again.
 All of our data is GONE! |
|
Quoted:
You know what, I'm just going to go ahead and post this video again.  https://www.youtube.com/watch?v=gSrnXgAmK8kSo much fail in this video. A good reason to not just use bullshit cobbled together because you thought it looked cool on a gaming website. |
|
Quoted:
So much fail in this video. A good reason to not just use bullshit cobbled together because you thought it looked cool on a gaming website. More like if you say your going to build an offsite backup then actually build and offsite backup. Otherwise be prepared to grab ankles. |
|
It usually starts off with the best of intentions. Then it is discovered just how involved and expensive it actually IS to do this correctly. That's when the duct tape and string comes out and things go downhill.
Quoted:
More like if you say your going to build an offsite backup then actually build and offsite backup. Otherwise be prepared to grab ankles. Quoted:
Quoted:
So much fail in this video. A good reason to not just use bullshit cobbled together because you thought it looked cool on a gaming website. More like if you say your going to build an offsite backup then actually build and offsite backup. Otherwise be prepared to grab ankles. |
|
so you want a metro cluster. N+1, windows uses something similar in WCS now plus a witness so you could probably make that work. Where are you backing up this data?
This is an async layout were talking about, ie if you lose one you could lose data. Generally worth the risk for a file server and a few minutes lag. EDIT: you are discussing an HA layout, fault tolerance. Where is this data copied and stored in some type of backup rotation. |
|
I still think it's hilarious that kid built a physical file server with 300,000 IOPS for basic SMB shares with mapped network drives for 6 users. I get using components you got for free from corporate sponsors, but come on.
Quoted:
It usually starts off with the best of intentions. Then it is discovered just how involved and expensive it actually IS to do this correctly. That's when the duct tape and string comes out and things go downhill. |
|
Quoted:
I still think it's hilarious that kid built a physical file server with 300,000 IOPS for basic SMB shares with mapped network drives for 6 users. I get using components you got for free from corporate sponsors, but come on. Didn't see his clients wired up with 4cards / 32GB FC to mount luns switches or Arista 100GB and NFS and the requisite cards to drive that IO. SMB falls apart well before that, with 6 users. I mean unless they were all using USC C460's it's all flash. |
Yep, this is the "computer guy" gold standard.
There are surprising numbers of "that guy" who thinks that the latest parts build in "PC Gamer" or "Maximum PC" is obviously kick-ass for business. After crawling forums looking for more "performance" he finds out about pedestrian but impressive-sounding things like SAS, RAID controllers, etc. and a new threat to critical business data everywhere is born. Businesses wind up hiring these guys because they find out how expensive it is to engage a professional services company or to hire an individual who actually knows what he/she is doing and will work for a small/medium business. It's an expensive lesson to have to learn. I always forget about that video but it's still funny every time I watch it. Quoted:
I still think it's hilarious that kid built a physical file server with 300,000 IOPS for basic SMB shares with mapped network drives for 6 users. I get using components you got for free from corporate sponsors, but come on. |
|
I had a customer once. They were a small CLEC, and the owners were basically just run-of-the-mill geeks that knew enough about networking get by building a WISP and riding the CLEC bandwagon in the mid-90s selling basic POTS and shitty DSL in a rural county.
They had VMware problems, they had vSphere hosts with 128GB of RAM, dual socket 16 core/32thread chips running over 50 VMs each, backed by 10 year old Thecus NAS boxes with 7200 RPM SATA drives in RAID 10. Shit was so deep into spin-lock that I was seeing storage latency spikes of over 2500ms, and average continuous latency of 450-500ms. Round and round I went with them, saying they needed to upgrade their storage. "We aren't a big company, we can't afford that." but during our initial discussion they loved to brag about their personal workstations with i7 processors, SSD/PCIe SSDs and water cooling. Quoted:
Yep, this is the "computer guy" gold standard.
There are surprising numbers of "that guy" who thinks that the latest parts build in "PC Gamer" or "Maximum PC" is obviously kick-ass for business. After crawling forums looking for more "performance" he finds out about pedestrian but impressive-sounding things like SAS, RAID controllers, etc. and a new threat to critical business data everywhere is born. Businesses wind up hiring these guys because they find out how expensive it is to engage a professional services company or to hire an individual who actually knows what he/she is doing and will work for a small/medium business. It's an expensive lesson to have to learn. I always forget about that video but it's still funny every time I watch it. |
|
Quoted:
Didn't see his clients wired up with 4cards / 32GB FC to mount luns switches or Arista 100GB and NFS and the requisite cards to drive that IO. SMB falls apart well before that, with 6 users. I mean unless they were all using USC C460's it's all flash. When that kid built that thing he probably sat around for a weekend wondering why no matter what he did, he couldn't it to go any faster than 110MB/s across the network. |
Dear God.
I would have backed out of the room slowly with my hands up going "yeah, you know, you guys have totally got this one by the balls and at this point I'm just an unnecessary expense... Maybe start running some of the more critical VMs on your kick-ass workstations! Good luck!" Quoted:
I had a customer once. They were a small CLEC, and the owners were basically just run-of-the-mill geeks that knew enough about networking get by building a WISP and riding the CLEC bandwagon in the mid-90s selling basic POTS and shitty DSL in a rural county. They had VMware problems, they had vSphere hosts with 128GB of RAM, dual socket 16 core/32thread chips running over 50 VMs each, backed by 10 year old Thecus NAS boxes with 7200 RPM SATA drives in RAID 10. Shit was so deep into spin-lock that I was seeing storage latency spikes of over 2500ms, and average continuous latency of 450-500ms. Round and round I went with them, saying they needed to upgrade their storage. "We aren't a big company, we can't afford that." but during our initial discussion they loved to brag about their personal workstations with i7 processors, SSD/PCIe SSDs and water cooling. Quoted:
I had a customer once. They were a small CLEC, and the owners were basically just run-of-the-mill geeks that knew enough about networking get by building a WISP and riding the CLEC bandwagon in the mid-90s selling basic POTS and shitty DSL in a rural county. They had VMware problems, they had vSphere hosts with 128GB of RAM, dual socket 16 core/32thread chips running over 50 VMs each, backed by 10 year old Thecus NAS boxes with 7200 RPM SATA drives in RAID 10. Shit was so deep into spin-lock that I was seeing storage latency spikes of over 2500ms, and average continuous latency of 450-500ms. Round and round I went with them, saying they needed to upgrade their storage. "We aren't a big company, we can't afford that." but during our initial discussion they loved to brag about their personal workstations with i7 processors, SSD/PCIe SSDs and water cooling. Quoted:
Yep, this is the "computer guy" gold standard.
There are surprising numbers of "that guy" who thinks that the latest parts build in "PC Gamer" or "Maximum PC" is obviously kick-ass for business. After crawling forums looking for more "performance" he finds out about pedestrian but impressive-sounding things like SAS, RAID controllers, etc. and a new threat to critical business data everywhere is born. Businesses wind up hiring these guys because they find out how expensive it is to engage a professional services company or to hire an individual who actually knows what he/she is doing and will work for a small/medium business. It's an expensive lesson to have to learn. I always forget about that video but it's still funny every time I watch it. |
|
That's pretty much how it went.
OH!! Another fun one. They had absolutely critical billing functions (moving of call detail records from their 15 year old CopperComm carrier switch into their billing system) running on ColdFusion scripts that were executed by Windows Scheduled Tasks on a 15 year old Windows XP physical workstation that just sat on a shelf in the server room. Quoted:
Dear God.
I would have backed out of the room slowly with my hands up going "yeah, you know, you guys have totally got this one by the balls and at this point I'm just an unnecessary expense... Maybe start running some of the more critical VMs on your kick-ass workstations! Good luck!" |
|
Quoted:
So what I want to know is why is this second WSS off site, but there's no replication in place? What's it doing off site? I put the second server offsite (we have an offsite Datacenter) in case of physical disaster at the first site (i.e. water main breaks over the server room here). This allows us to have a complete copy of our data that we can restore from. Server number one does have some fault tolerance. The main system drive (C:) is mirrored. So if the main drive #1 fails it will switch over to the #2 drive and keep going. The storage RAID 5 array which includes a standby hot swap disk, so if we lose a drive it will keep going seamlessly. |
|
So there is replication between WSS A and WSS B?
So, if you want seamless fail-over for User directories, you can totally do this with DFS-R. The one caveat is that if there are shares that multiple users have access to, with DFS-R you can't have file locking unless you install a 3rd party file locking service. This means if more than one user opens the same file, shit gets weird. Quoted:
I put the second server offsite (we have an offsite Datacenter) in case of physical disaster at the first site (i.e. water main breaks over the server room here). This allows us to have a complete copy of our data that we can restore from. Server number one does have some fault tolerance. The main system drive (C:) is mirrored. So if the main drive #1 fails it will switch over to the #2 drive and keep going. The storage RAID 5 array which includes a standby hot swap disk, so if we lose a drive it will keep going seamlessly. |
|
Quoted:
It usually starts off with the best of intentions. Then it is discovered just how involved and expensive it actually IS to do this correctly. That's when the duct tape and string comes out and things go downhill. Yeah those good intentions go right down the drain when your data is hosed and you gotta dish out big money for data recovery. |
|
Quoted:
So there is replication between WSS A and WSS B? So, if you want seamless fail-over for User directories, you can totally do this with DFS-R. The one caveat is that if there are shares that multiple users have access to, with DFS-R you can't have file locking unless you install a 3rd party file locking service. This means if more than one user opens the same file, shit gets weird. Crap. I do have shares that multiple users have access to. No, there's no replication between WSS A and WSS B. I could not get replication to work when I tried configuring it (using the "Replication Group Wizard"). Kept getting an error that told me my replication group could not be created. Only the Domain Admin of our whole University could set up the new group for me which is a real hassle. *I also heard that replication of .pst files (for Outlook) cannot be done if the .pst file is in use. Which, if true, is the nail in the coffin because we use .pst's extensively. Currently I'm only using the Windows Backup that came with 2012 to backup WSS A --> WSS B. |
|
If you're using Exchange, then PSTs can be safely ignored/filtered from replication as all of that data is server side. If you use PST files extensively, and aren't using Exchange, then you're environment is fucked up, your management team is fucked up, and you should just quietly resign yourself to the shitstorm that will eventually explode all over the place. So if you can't get replication working because your change management is a "hassle" why the hell is that WSS there? Who approved the purchase but didn't approve the change requirements to actually deploy it?
Quoted:
Crap. I do have shares that multiple users have access to. No, there's no replication between WSS A and WSS B. I could not get replication to work when I tried configuring it (using the "Replication Group Wizard"). Kept getting an error that told me my replication group could not be created. Only the Domain Admin of our whole University could set up the new group for me which is a real hassle. *I also heard that replication of .pst files (for Outlook) cannot be done if the .pst file is in use. Which, if true, is the nail in the coffin because we use .pst's extensively. Currently I'm only using the Windows Backup that came with 2012 to backup WSS A --> WSS B. |