If you don't have Solarwinds HA: * Why not?* Cost? * Your company doesn't believe monitoring is important enough to support (with licenses, employee setup/support hours, hardware environment, etc.)? * You've never thought about it? * You just don't need it? * You don't have time to set it up or to maintain it? If you DO have Solarwinds HA: * Why did you get it? * How did you convince your company it is necessary? * How satisfied are you with it? * Have you set up HA for ALL your polling engines? * What would you change about it? * Have you seen the new HA administration view that shows all your standby / HA pollers, and that highlights any differences between them? You can check out some of the HA views on your own Main Poller here: * https:// /ui/ha/settings * https:// /ui/ha/summary

Community
- Command Central
- MVP Program
- Monthly Mission
- Blogs
- Groups
- Events
- Media Vault
Products
- Observability
- Network Management
- Application Management
- IT Security
- IT Service Management
- System Management
- Database Management
Content Exchange
- SolarWinds Platform
- Server & Application Monitor
- Database Performance Analyzer
- Server Configuration Monitor
- Network Performance Monitor
- Network Configuration Manager
- SQL Sentry
- Web Help Desk
Free Tools & Trials
Store

Let's talk about Solarwinds High Availability. Do you want HA? Why? Do you have it? Why'd you get it? Does it work well?

rschroeder

If you don't have Solarwinds HA:

Why not?
- Cost?
- Your company doesn't believe monitoring is important enough to support (with licenses, employee setup/support hours, hardware environment, etc.)?
- You've never thought about it?
- You just don't need it?
- You don't have time to set it up or to maintain it?

If you DO have Solarwinds HA:

Why did you get it?
How did you convince your company it is necessary?
How satisfied are you with it?
Have you set up HA for ALL your polling engines?
What would you change about it?
Have you seen the new HA administration view that shows all your standby / HA pollers, and that highlights any differences between them?

You can check out some of the HA views on your own Main Poller here:

https://<YOUR SOLARWINDS MAIN POLLER ADDRESS OR DNS>/ui/ha/settings
https://<YOUR SOLARWINDS MAIN POLLER ADDRESS OR DNS>/ui/ha/summary

Find more posts tagged with

Accepted answers

All comments

rschroeder

NO one has SW HA? Seriously?

aLTeReGo

There are actually a lot of people who own HA. In fact, more than owned FoE. This just doesn't happen to be a very active forum because HA simply works. if you're wanting to reach a broader audience to improve your chances of someone running hA seeing your post, and willing to reply, you might want to try the Network Performance Monitor forum.

CourtesyIT

I have concerns about the operations due to some of my low-bandwidth/High Latent circuits from US to Euorpe/Pacific.

I also have concerns about the SQL Database portion of HA.

Eric

rschroeder

Thank you!

I was hesitant to put this there, since I've had helpful Thwack admins move my content to more specific forums in the past. But I'll give it a shot.

jhandberg

We have talked about it since it's introduction and really want to do it. Budgeting is the issue here right now. Our monitoring group doesn't have their own official budget and no one has fully figured out where our budget is ultimately going to come from.

aLTeReGo

Orion High Availability will work better than an APE across low bandwidth, highly laitant links. The database replication is handled natively by SQL's AlwaysOn Availability groups which can be used in a multi-subnet failover configuration if so desired. .

rschroeder

jhandberg I was in that same boat until I discovered NAM licensing. I pay less now for SW support than I did when I had stand-alone versions of NPM, NTA, NCM, and three APE's. And I got UDT, IPAM, VNQM, HA, and can install up to 20 APE's at no additional cost.

Last year I knew I I needed more polling engines.

I also needed UDT for our CMDB.

And I needed VNQM for my Telecom team.

And HA would benefit my team and all our SW customers..

And I wanted to learn more about SW IPAM (I already have Infoblox DNS/DHCP/IPAM, and knew there'd be little budget available to duplicate part of it).

I learned about NAM licensing almost a year ago, and had prepped my IT Director for considering it with the 2018 budget cycle. I explained how UDT would provide the switch/blade/port info for every device, and that our CMDB could mine the Solarwinds Orion SQL database and extract that information to our LAN Desk solution. That UDT would also save the Network team from continually having to go through a tedious process to discover the port any device is plugged into (ping the device's address, SSH to the router in that subnet, sho ip arp | in that address, then SSH through a Distribution switch and possibly several Access Switches to the switch). Instead we just copy & paste the IP address or MAC address of the client, and the switch/blade/port it's attached to shows up.

With that prep in place, last fall I reached out to my SW SW for a quote. It came back very high, and I said thanks, but I can't bring that to my boss's boss for something unbudgeted. The SE hemmed & hawed a while, then noted I was Solarwinds Certified, which I verified. He hemmed & hawed some more, and then discovered I'm a Thwack MVP, and I confirmed that, too. He trimmed double-digit percentages off the original quote, and I we discussed the ramifications it would have on the annual support contract. After the initial up-front buy-in, the annual support contract went down, and I ended up with the ability to install up to 20 pollers and monitor up to 100,000 elements. I brought the quote to the IT Director, reminded him of the project and its benefits, mentioned buying this would decrease our annual Support Contract costs, and went home for the day.

The following morning I found the electronic invoice and the Activation Codes in my Inbox! My Director had found the funds and made it happen.

The great challenge in IT is always finding time to install and learn what you ask for, but I'd have a pretty good time of it. I've installed more pollers, IPAM, UDT, and VNQM, and am still waiting to deploy HA. It's why I created this post.

It's not hard to see HA would solve some issues of various teams not having access to NPM during its upgrades / reboots / hotfixes. Duplicating all my Pollers is simple--they already run on VMs. I just haven't asked for the duplicates yet, since I'm in the middle of some bigger projects (replacing core and distribution hardware in my main campus, and replacing 400 Cisco 2960S's with 3850's for ISE compatibility).

Your goal is to find the budget. You probably read the paper I wrote a few years about that topic. It still applies: A Stratagem For Obtaining Funding For Your Projects

Play the game that the decision-makers and budget stake holders play. Once you find the right buttons to push, and can couch your request in the proper context, using their favorite buzz words, you stand a chance at getting the funds.

Good luck!

Rick

jhandberg

When you first posted about your NAM purchase, I forwarded that to my supervisor and he started talking to our sales rep. None of us are MVPs, so that never came up. The price quote was significantly higher, no 40% discount. We also regularly use SAM, WPM, SRM, and VMAN, which were not officially supported on NAM. We do have 9 APE's and are around 2-3k short of the recommended max elements. EDIT: just under 20k elements short, but still....

All this monitoring and proactive alerts has become the hot topic, but we have no official budget. We have to beg and borrow from other silo budgets. Network is paying part of the maintenance, as is the server group. We don't have an official charge-back to the departments we monitor devices in. I think my paycheck even comes from some other agency's budget. They have talked about sorting all that out since I started here 1 3/4 years ago. I just shrug my shoulders (state government, what are you going to do) and keep putting out my recommendations. I will keep recommending HA and maybe NOM or NAM.

Lately my supervisor and I were talking about what we are going to do when we hit that magical 100,000 element max. Maybe split the networking off into another instance? Then pull it in under the EOC if that works right? We are not sure, but we have many more agencies to centralize under our monitoring umbrella yet. That might be on opportunity to rethink licensing.

rschroeder

My organization merged with two others a few years ago; each had their own NPM solution in place, so we bought EOC to give visibility into all of them.

That was a mistake.

It segmented the Network "Team" and facilitated regional members to focus solely on "their" sites, and not help out on other region's sites.

I worked a deal with SW to get rid of EOC and convert two Main Polling Instances to APE's, then report everything to one Main Poller. And just like that, the team was back together, working on every site--simply because they all showed up in the same pane of glass. Better still, it turns out APE's were less expensive than running three separate main NPM instances, so licensing/support costs dropped when I dropped EOC and converted two of those three servers to APE's.

I didn't like EOC because it was much slower at displaying issues than the individual Main Polling instances. But I can see where it fits in, for organizations that exceed that 100K limit.

For my $.02, I'd rather start up a completely separate NPM instance and begin growing it out to its own 100K elements, and have two screens showing both NPM's, than use EOC. Unless EOC has become more affordable and has much-improved performance. Then I might look into it when my organization's monitoring needs have grown that much.

jhandberg

We have had the EOC for a long time, since before I started here. Even the newest version needs improvement. We have tried to create custom views in it for others, but when others log in, they can't see the views. Something isn't right there, and SolarWinds support hasn't figured out what. It isn't the ideal solution right now, but we have been told by support that our database tables are very large and may be causing some issues. We started shortening retention levels to compensate, but they are still large, and we have less historical detail.

If we split off networking, we would kind of silo that, but it might make other module configurations available between instances. I do wonder about something like site dependencies with mixed network/server/application infrastructure in that scenario. The other thought was split regional. Cut the state in half or something. Time will tell.

Budget issues need to be resolved first though. Ok, that and the state legislature approving my employment contract. Kind of working without one now. We might be working under the old that expired last July 1st. Yes, contract and budget need to be the priority.

RichardLetts

Honestly:

Strength:

we have a well-working FoE install.

Weakness:

My Orion Admin quit at the end of the year (I'm hiring if anyone is interested).

The server operating systems we are running on are not supported by 12.2 and WAN-HA in that version

I am being forced off FoE in March (and not enough time to get the changes through to move to LAN-HA) in time

Opportunity:

We need to do a server refresh in the next FY as the warranty expires in July which will give us a platform we can run WAN-HA on.

Threat:

Our FoE licenses will stop working in May when they are up for renewal in any case

SolarWinds has failed to deliver key improvements applicable to our environment in a number of products over the past 7 years. among these are SAML so we can have two-factor authentication, VRF polling for Juniper in UDT, and the ability to edit the Vendor and MachineTypes. [we've stopped using other product I had requested enhancement for including IPAM, and NTA]

ryhlen.schoeberl

We Have HA. It does suck changing IP addresses when switching from no HA to HA. but it is great for if i have issues with my main server and need to reboot the system that our NOC can still monitor things with the other server.

We bought NAM and it came with alot of the features we were already using ourt polling was at 80% after i adjusted polling times. With NAM you get NCM, NTA, IPAM, UDT, NPM, QOE, 5 polling licenses, HA. To convince them a getting NAM for 2000 nodes was just a little more money than purchasing 1 additional polling server. As well as renewal costs are quite reasonable.

I am really satisfied with HA. I wish we would have would have done it at the start.

jm_sysadmin

I bought HA this morning, I will let you know how it goes.

Basically it boiled down to IS not wanting to lose visibility into its systems in the event of failures. We run much of the stack, NPM, NCM, NTA, SAM, SRM, IPAM, DPA, and VMAN. Losing all of them at once was not OK. We had a small fire in our primary data center recently, Orion was up the whole time, but had it been gone, we would not have understood the scope of the issues and how to prioritize work. HA discussions started shortly after. If you want to know more about my roll out as I go, I'd be happy to talk about it. Just send me a message and I will keep you up to date.

grizzlyferrett

Why did you get it?
- Purchased at time we purchased other SolarWinds products and FoE was no longer available.
How did you convince your company it is necessary?
- We needed a platform that was always on so it made sense to have HA. It works well for Windows Updates and we can re-start servers without affecting monitoring or access to the Web portal.
How satisfied are you with it?
- Very satisfied
Have you set up HA for ALL your polling engines?
- Yes
What would you change about it?
- I would like to the platform to be available even during upgrades of the application. Having to take SolarWinds off-line to upgrade is not very Highly Available.
Have you seen the new HA administration view that shows all your standby / HA pollers, and that highlights any differences between them?
- No, what is this?

yaquaholic

Why did you get it?

We run monitoring as a service and when people are paying for something, they tend to get a little upset when it's not available.

How did you convince your company it is necessary?

Was within the requirements for the monitoring system and the budget was much less of a concern back then.

How satisfied are you with it?

Most of the time, very happy.

We did have a few delayed fail-overs (8 minutes) and triggering hundreds of false alerts, never resolved; though thankfully never re-encountered.

And product upgrades, large lumps of downtime, not good.

Have you set up HA for ALL your polling engines?

No only the core Orion, the APEs and AWS are not under HA (budget).

What would you change about it?

HA should allow for patching, hot-fixing and upgrading. Every upgrade leaves me with hours of downtime and missing data.

We run another basic monitoring system just for such times, so we aren't completely blind, but those gaps in the data in the reports are still really obvious to the customers.

Have you seen the new HA administration view that shows all your standby / HA pollers, and that highlights any differences between them?

If you are referring to http://<orion>/ui/ha/summary, then yes I have.

I could with a reminder on the CLI controlling of HA though, for those times the GUI is not available.

mesverrum

Ignore this if this doesn't apply to you, but whenever I see orgs who are maxing out the 100k limit I have to ask if there is actual material value in monitoring bandwidth on all those individual interfaces?

A big UDT license would give you up/down/connected device visibility into all the switches, and is a hell of a lot cheaper than standing up a completely second instance/EOC/etc. Does every endpoint really need that bandwidth history recorded? I figure that it makes sense to use NPM licenses/resources to monitor everything that is up on the core and distribution layers, but only infrastructure uplinks at the edges. Remove unused interfaces from NPM until they show as up in a discovery or via UDT.

Are you monitoring physical server uplinks from the server OS and the same interface from the switch perspective? If things are getting tight wouldn't it make sense to consider dropping the interface monitoring at one side or the other?

Have you ever tried to run a report to see how many interfaces you are monitoring that have not passed traffic in over 24 hours? I bet you would find a lot of things in there that maybe aren't getting a business ROI from monitoring if you really dug into it.

I've had to make reports for clients where we itemized the licensing and ongoing maintenance costs associated with monitoring every object in the environment so they could use that as part of the bill back model they were setting up. Sometimes when you see it in black and red it raises the question of "is this node really worth $345 of ongoing maintenance support"

I have never found an instance that I couldn't trim a huge percent of their interface count away without having any negative impact on their actual visibility into problems. Usually increases admin overhead to keep track of all these things instead of just blanket monitoring everything, but with good use of custom reports and the API you can cut down the admin burden by a mountain. Even if you need to book training and pro services to get you over that hump the ROI on training tends to be a lot better than standing up a second environment if you could have avoided it.

This is also somewhat off track from the HA topic though. I've installed it many places and like aLTeReGo said it is pretty simple and just works where I've used it. One thing I wanted to add though, you mentioned seeing HA as a solution during upgrades. Unfortunately as it exists right now you still don't have the capability to do an upgrade without taking the system down, I didn't want anyone to get their hopes up. Seems like that would be tricky since some upgrades make structural changes to the database and I expect it would be pretty hard to keep the system hot while you did those.

Along the lines of some of the points I made above, HA also needs to be looked at in terms of ROI and impact. Many orgs don't even allocate a proper budget to do monitoring, let alone setting aside additional money to offset a few hours a year of downtime. Downtime of your monitoring tool is not downtime of your environment (unless maybe you run an MSP or similar) so the value for HA is in the potential it offers you to be notified of a problem during those windows when your system would otherwise have been down. If you add up the dollars and cents that is often such an edge case scenario that many people just figure they can accept the risk and move on without it. I am sure the business impact within a hospital like your environment rschroeder is a lot more dire than it would be for some other types of business so I can definitely appreciate the use case there.

-Marc Netterfield

Loop1 Systems: SolarWinds Training and Professional Services

LinkedIN: Loop1 Systems
Facebook: Loop1 Systems
Twitter: @Loop1Systems

rschroeder

You can find that view starting here: Settings > All Settings >

HerrDoktor

i do have 2 clients with HA

Why did you get it?

Requirement of 24/7 Monitoring. One of those two clients is reporting SLAs for their customers and need the system to run all the time

How did you convince your company it is necessary?

Management set the requirement

How satisfied are you with it?

Meh....HA broke 3 times during a failover test within the last 6 months

Have you set up HA for ALL your polling engines?

YES all polling engines

What would you change about it?

ability to do updates of solarwinds without downtime

make it more robust

Have you seen the new HA administration view that shows all your standby / HA pollers, and that highlights any differences between them?

yes

david.botfield

We have it but I'm not involved in the day to day running of the server. I just use the Solarwinds elements. We have lots of our own plus customer devices monitored and we report from Solarwinds on this. so the need for high availabilty is quite pressing. We're not on the latest versions at the moment (due this month or next) so I haven't seen the latest in this area.

rschroeder

There's been so MANY great comments and reports about HA in this thread--THANK YOU all!

The biggest red flag to me is HA doesn't keep the monitoring & reporting going during maintenance windows. My original intent was to ensure we had Solarwinds network status reporting 7x24, no matter the maintenance.

I'd very much like this to be addressed. And until it's corrected, the HA Marketing info should include that caveat up front, and list exactly what HA can NOT do to ensure 7x24 monitoring and reporting goes on.

Right next to that, I'd like to see what issues HA addresses, how it makes things keep working--except for those specific conditions when it DOESN'T keep reporting & monitoring going.

[Deleted User]

We don't have HA - A bone of contention for me and something I will address in time.

Why not?

- Cost - This will be one of the issues, of course. However, the costs of failure are much greater! (In my opinion!)

- You don't have time to set it up or to maintain it? - From what I've read so far, it doesn't actually look particularly labour intensive in this regard!

jlhartsock

If you don't have Solarwinds HA:

Why not?The lack of support for zero downtime upgrades turned this effort into a non-starter for the conversation within my company.
- Cost? The cost could be easily justified if we did not incur outages during upgrades. However, since that wasn't on the table, I couldn't provide a justification for the additional cost.
- Your company doesn't believe monitoring is important enough to support (with licenses, employee setup/support hours, hardware environment, etc.)? See statement above.
- You've never thought about it? I was extremely excited when I learned that SolarWinds had an HA option after I had taken over the platform last year. This was quickly extinguished when I met with the sales team and found that we would still experience the 4+ hours outages for upgrades.

designerfx

jlhartsock most upgrades now range in the 30m-1h ish, primary or APE. They have a new installer and it's beyond significantly faster. serena can probably vouch for how much the installer has made a difference but I don't know if she can divulge average install time improvements as percentage.

jm_sysadmin

Agree, a 4+ hour upgrade with multiple products installed a year ago was expected. Now with most of the Orion products and their hotfixes included in a single run, My actual time down was under 30 minutes.

jlhartsock

I'm going to perform another batch of upgrades within a month so hopefully I see the same experience. My last upgrade in October ran just under 4 hours when running the installer and config wizard on my primary server and 3 APEs.

aLTeReGo

Note that you can run the Configuration Wizard on all your Additional Polling Engines in parallel now, so there's no need to perform this function serially. Unless of course you also enjoy watching paint dry.

designerfx

jlhartsock keep in mind the *APE* installer has changed if you're not on at least NPM 12.01. See: NPM 12.0.1 Release Notes - SolarWinds Worldwide, LLC. Help and Support . First you upgrade your primary and then you can download the new APE which is a smart installer and basically handles the rest. If you haven't upgraded in four months I'm wondering how far back your updates are. That might make things take longer.

rschroeder

I foresee a positive experience, assuming your versions are not too many generations behind. The latest installers / hotfixes are pretty good stuff.

rschroeder

I was as 12.0.1 in September; when upgrading to 12.2, I found doing four APE's simultaneously caused problems. It was better to let the first APE complete downloading its Polling Engine files/updates from the Main Poller before starting a second APE, or a third or fourth.

Now, with much-improved updaters, I think the whole process must be much improved.

jlhartsock

rschroeder Sweet! I'm looking forward to it. We're not far behind so my expectations are a bit uplifted from the last experience.

Orion Platform 2017.3, QoE 2.4, CloudMonitoring 1.0.0, SAM 6.4.0, NPM 12.2, IPAM 4.5.2, SRM 6.5.0, NTA 4.2.3, VIM 8.0.0, DPAIM 11.0.0, NCM 7.7, NetPath 1.1.2

aLTeReGo HAHAHA! Definitely took advantage of running them in parallel the last go around. That would be mind numbing to have to run them in serial.

jhandberg

I agree on that point. We have experienced troubles starting the process on all APE's at ones. But once the files are downloaded one at a time, they seem to all run.

hpstech

Bump for more experience and feedback in the field with HA.

Thanks

wlouisharris

I found this post doing some research on HA. We implemented HA in QA and PR. Here is my feedback:

Why did you get it? Our customer wanted high availability.
- We had skepticism but the customer wanted to be able to check off the box to say their services are redundant.
How did you convince your company it is necessary?
- Same as above, customer demand.
How satisfied are you with it?
- So far the HA service works pretty good using BIND DNS. We also have SQL HA. We had an issue where both servers went into standby mode and we had to put in a support call to get the main poller back in active mode.
- The best use of HA is for Microsoft/operating system patching. We can move servers to standby while we patch.
- The other use case I can think of for HA is if one of the main pollers has some type of o/s corruption. Having a standby can make this seamless. We had one case with NPM 12.0 where the main poller became corrupt, but with Windows Server 2016 and NPM 12.2 we have seen a lot more stability. Even so it's an advantage.
- Make sure you implement this in a lower life cycle that closely resembles your production environment. We had to purchase an SL100 license for QA; otherwise you only get a 30 day evaluation. It's important for us to do our upgrades in QA first prior to PR so we need a persistent QA environment.
Have you set up HA for ALL your polling engines?
- Yes
What would you change about it?
- This product does not offer HA during Solarwinds upgrades or patches. This is a huge drawback.
- We would like a better DR recovery process. We would like to have the ability to bring a 3rd server online and replace this as a standby or active primary poller; then remove the original server.
Have you seen the new HA administration view that shows all your standby / HA pollers, and that highlights any differences between them?
- no

ecklerwr1

I got rid of my 2960's and have many 3850's but the real magic seems to come with 9300's. We've swapped all our IDF closet switches for 9300 stacks now. I'm working on another smaller air gapped network and my goal is to go all 9000's now. I also had FoE for years and it was total waste of money. I suppose I'll follow this to see how your HA goes. ;^}

rschroeder

No one here is willing to claim NMS is important enough to require resources for HA. I have the HA license, haven't implemented it due to lack of resource allocation. Sigh . . .

Just watch them try to respond in timely manner if/when our NPM and NCM is down long-term for any reason. Suddenly the users will become the Alerts, and the Help Desk will be flooded longer and longer.

aLTeReGo

Native SAML 2.0 support for Orion is included in the NPM 12.4 Release Candidate which can be downloaded from your Customer Portal now.

rschroeder

I'm just embarking on a forklift replacement of about a hundred 2960S stacked switches with 9300's. Some of us are excited about the virtual drive space allocated in the 9000's. Others of us have seen such resources abused by malware or hackers in earlier systems, or are frustrated by the lack of funding for formal training to ensure we use the new systems the best way possible. Time will tell how I feel about it. Certainly it represents opportunities for more granular work, more automated work, and more one-off configurations. That doesn't equate to simpler and more deployments that conform 100% to a baseline.

gregrybak

wlouisharri,sI see you use BIND DNS, what is your DNS system (I'm hoping you'd say Infoblox)?

aLTeReGo

Many customers use Infoblox with Orion HA.

wlouisharris

We use Bluecat DNS. We had to end up using a 128 bit MD5 key to get it to work. Of course the newer security standard is SHA2(256 bit)keys. Basically we had our DNS team create a new zone for Solarwinds called solarwinds-ha; then they had to delegate us to have root privileges within that zone. From a process standpoint it was easier for us to use BIND DNS, but most of the internal Solarwinds support documentation is written with Active Directory DNS in mind. We spent quite a bit of time with support, our DNS team, and trial and error in QA to get it to work.

Right now we use HA more like disaster recovery which is really quite an expensive DR solution.

Best of luck.