Highlighted
Trusted Contributor.
Trusted Contributor.
1322 views

NA Multimaster vs. Horizontal Scalability for HA Environment

I would like to implement the best solution and although the recommended solution is to deploy Horizontal Scalability, I only have 2 Windows servers each with MS SQL installed on them. Currently we're only supporting <250 devices in a highly available, highly sensitive environment with low latency WAN.

Thus, I think Multimaster is the way to go but I want to get any and all feedback on this if people have some history with running two redundant master cores with local SQL databases.

Any advice?

0 Likes
7 Replies
Highlighted
Acclaimed Contributor.
Acclaimed Contributor.

Re: NA Multimaster vs. Horizontal Scalability for HA Environment

Hi,

 

I've not been a personal fan of MM, I'm more a fan of HS.  Just less moving bits and overhead.  Elsewhere I had a MM two core NA system and it was problematic.  Though I've done similar with SQL Server and it's been OK.  Not really a fan of local DBs, it works, but just prefer separation of duties for production.  

With MM, you just have more overhead, sure it's not as much as if you had 3 or 4 cores, but still there.  

But I'd want to know more than you have provided, such as:

 

Are your cores located at same location or separate locations?

The two cores, do you plan to have both active or is this more a standby scenario? 

What's the key - raw performance or availablity?  If it's raw performance, I'd say look at Linux, though only 250 devices, less of issue.  

There's no mention of hardware - CPU, memory for example - again, if this is about performance, these items are key.  

-Chris

0 Likes
Highlighted
Trusted Contributor.
Trusted Contributor.

Re: NA Multimaster vs. Horizontal Scalability for HA Environment

Hey Chris,

The key here is definitely availability of the information. Obviously performance is nice and I don't think we'll have issues since we don't have a lot of nodes to manage.

Re: the cores, they're located in separate data centers. All servers are HP Proliant DL385p G8, Windows Server 2008 R2, 32 GB memory and AMD Opteron 6320, 8 CPU clocked at 2.80 Ghz with a license of MS SQL.

It makes more sense to be able to have a DB shared by two cores with a redundant DB on standby, which is HS, will it work having the primary DB on the primary core? If the primary core server fails, will the #2 core and DB take over?  I understand separation of duties but I'm only given 2 servers to work with in the environment.

What do you feel is best in this scenario?

Thanks again!

0 Likes
Highlighted
Acclaimed Contributor.
Acclaimed Contributor.

Re: NA Multimaster vs. Horizontal Scalability for HA Environment

OK, I think I understand better now.

Maybe take a look at the DR option, it'll definitely handle your two cores and provide high availablity.  

I still would prefer you split off the DB from App server, but understand you have the hardware you have.  

But this would allow you to set up DB relplication between the two SQL Server instances and it'll pass data back - forth.

You tag one core as active and the other as inactive.  

This will give you a good option for utilizing two cores in different locations.  

Maybe later you could add more hardware and have (say) two cores at each location and you'll separate the DB.  

This would provide you with great performance and availability.  

Let me know if you have any questions.

-Chris

 

 

Highlighted
Trusted Contributor.
Trusted Contributor.

Re: NA Multimaster vs. Horizontal Scalability for HA Environment

Chris,

Thanks so much for that documentation, it gives me some insight into a different approach. I am conflicted - I like the approach represented in appendix A of the NA DR Setup, but since database failovers require heavy manual intervention, the Multimaster approach seems to be the most seemless high availability solution. 

Here is my deduction to Multimaster after reading Horizontal, Multimaster and DR:
- No division of duty between core and DB for each datacenter location
- I am the primary SME of NA in my organization, meaning I'll have to recover the box if it fails. I want to employ the KISS method here since NA is a secondary duty, but I want NA to be in an HA pair
- Datacenter locations are separated geographically
- Latency between datacenters on this VRF/VPRN is <1ms
- If one datacenter/server dies, that core and database go too since they're one box. In this scenario, Chapter 4 of the Multimaster guide explains how to recover if one side dies. The only manual intervention needed here is to turn off replication on the subscriber if the publisher fails. Obvious there're some steps if the subscriber fails, but nothing too bad. Would we be able to restore NA Core\Database using a server backup from the previous day and then use the non-affected database to restore?

Can you sell me on using the disaster recovery configuration a bit more or do I have my answer? I don't mind havine core 1 and core 2 using the active db on core 1, but how much effort does it take to make core 2 db become active? From my perspective, it seems pretty tedious to perform the steps in Chapter 3: Switchover and Chapter 4: Switchback of the NA Disaster Recovery guide in the event that something happens to one of the boxes. I may be reading it wrong and it may be simpler given that NA Core and DB are on one device in my scenario.

Thanks for your time, this is great.

0 Likes
Highlighted
Acclaimed Contributor.
Acclaimed Contributor.

Re: NA Multimaster vs. Horizontal Scalability for HA Environment

Sorry for the delay, totally missed your update.  

Couple things - I totally understand your comments about manual work being required.  There are a few options.

1) You could take advantage of something like HP OO to either toally or partially automate this.

2) If you don't have OO, you could do this with scripts and triggered either by something similar to OO or someone in NOC

3) If I recall, you had a limited environment - so let's say you go with this for a year, you'll have one or two NA updates (one this year and then next year - typically around HP DIscover) - who knows, there may be additional automation for this added in.  🙂  

If you do DR, you need to make sure that you have DB replication between the two DBs - unfortunately, that's not something I can help much with, other than to say you need to talk with a DBA.  

I'd personally be nervous with having Core 2 point to Core 1 DB over a WAN - you don't want a slow link, that just spells problems but sounds like you're set there.  You could "try" this and see how it goes for that year, planning on building your setup out.  🙂  

In a nutshell, DR works as this:

Core & DB at location 1 are active

Core & DB at location 2 are inactive / down - DB replication is happening in backgroud.  

Say location 1 has an issue, you bring Core 1 and DB 1 down, bring Core 2 and DB 2 up.  So, you'd need some DB help (or scripting / automation).  The NA side of things - it's really easy to do as long as you have server / proxy access.  

Even if you had two cores up, you'd have some work to do if say Core 1 dies as you'd want to make sure that Core2 knows Core 1 isn't able to participate.  

Again, sorry I missed your update.  

-Chris

 

0 Likes
Highlighted
Contributor.
Contributor.

Re: NA Multimaster vs. Horizontal Scalability for HA Environment

Hello,

 I have a similar situation in that I have 3 geographicly separate NA 9.2 servers on RH 5 Linux running Oracle 10g that are in MultiMaster and having issues with replication. Only 2 of the servers were doing replication and one acting as primary has started having issues with oracle by requiring a new system.dbf tablespace file daily. I'm planning a new install on VM's runing RH 7.3 using Oracle RAC with NA 10.3 plus 2 additional NA proxy servers at separate locations. UNtil then, what's the best way to correct the daily need for additional tablespace in Oracle. This system is one I inherited a year ago.  T

Thanks

0 Likes
Highlighted
Micro Focus Expert
Micro Focus Expert

Re: NA Multimaster vs. Horizontal Scalability for HA Environment

Hi @stocksj1,

for tablespace needs you can consult with your DBA team to set an auto extend to the tablespace or have a datafile appended to the tablespace to be able to write, regardless a MM is a high maintenance system, if you need to upgrade it has to be break to do so then rebuild. and as you might have notice NA has to do little jumps to upgrade no straight line method for an upgrade unfortunately.

Pedro A. Batista
Customer Support Engineer

If you find that this or any other post resolves your issue, please be sure to mark it as an accepted solution.
If you are satisfied with anyone’s response please remember to give them a KUDOS by clicking on the THUMB at the bottom left of the post and show your appreciation.
0 Likes
The opinions expressed above are the personal opinions of the authors, not of Micro Focus. By using this site, you accept the Terms of Use and Rules of Participation. Certain versions of content ("Material") accessible here may contain branding from Hewlett-Packard Company (now HP Inc.) and Hewlett Packard Enterprise Company. As of September 1, 2017, the Material is now offered by Micro Focus, a separately owned and operated company. Any reference to the HP and Hewlett Packard Enterprise/HPE marks is historical in nature, and the HP and Hewlett Packard Enterprise/HPE marks are the property of their respective owners.