9 tools to help you manage your pet servers

Containers, orchestration, automation, full remote management, cloud-style distributed app development, continuous developer-continuous deployment and devops are the wave of future IT, and what sophisticated IT departments should be aiming for in the future. But there’s almost certainly at least one server your business relies on that is a unique and special snowflake — a pet you’ve named, rather than just another head of cattle in your data center that you can easily replace if it ever fails. Many companies that say they have private cloud have far less automated virtualisation environments, so it might be more than one.

Whether it’s an elderly application you’re not ready to replace, a control system for hardware where the vendor has never come out with a newer version, or any other server that you can’t – or aren’t ready to – move to the new world, you have to keep it running smoothly while you plan for how you eventually replace it. Here are some tools to make that more bearable, and more reliable.

The first thing to avoid with a snowflake server is change, because change makes these kinds of custom systems fragile. ScriptRock’s GuardRail is a great visual tool for getting an at-a-glance view of any changes to the configuration or your pet server – or just seeing exactly how different it is from any other servers you’re looking at. If it’s a server you can reach on the network, you don’t even need to install an agent to scan it.

GuardRail shows the differences in settings (compared to another system or an earlier scan of the same server) visually as a chart or a wheel, split into the different features and services on the system, from Windows Server features and roles, to environment variables and hotfixes, to SQL schemas, to ports and installed hardware. That makes spotting what’s changed very fast; with additions, deletions and changes clearly color-coded. You can search by keyword or click to drill down into a specific area.

This is ideal for servers that you have to support, but where someone else might be making changes to the workload. You can also use it to make sure the DevTest environment has the same setup as the live server.

If you can’t easily rebuild or replace a server, you want to make sure it doesn’t fail. One option is adding redundant hardware to the system, but if you can virtualize the workload you have a lot more options. Go beyond running it on Hyper-V or VMware and look at whether you can run it on an IaaS public cloud or set up a real private cloud system. Unless your pet server is tied to specific hardware, you can take out a lot of the work by putting it on a truly automated, distributed system. You can build your own with the Windows Azure Pack (shortly to become the Microsoft Azure Stack and be much closer to Azure) or run on “hyperconverged infrastructure,” where you buy a preconfigured private cloud appliance like Microsoft’s Cloud Platform System (and equivalent systems from HP and Dell) or Simplivity’s OmniStack (which you can buy directly, or from Cisco or Lenovo). Whichever configuration you pick, it’ll let you stop worrying about the hardware and concentrate on keeping the workload running.

It’s critical that you have a backup for when the server hardware fails (which is as inevitable as ending up with workloads that you can’t simply replace with a new instance any time there’s a problem). You can stick with simple disk imaging tools, or use specific server backup tools like Acronis Backup Advanced, Paragon Hard Disk Manager – which uses both Microsoft VSS and Paragon Hot Backup to snapshot live files – or Macrium Reflect Server (which creates images you can convert to virtual machines or deploy to different hardware). You can back up SQL Server, Hyper-V VMs, SharePoint, Exchange and Windows clients to Azure Backup without needing System Center Data Protection Manager in your infrastructure, by using the new Microsoft Azure Backup Server (which can run locally or on Azure).

And if you can virtualize your server on Hyper-V or VMware, Veeam Backup and Replication gives you excellent backup management, very fast recovery and the option to send snapshots directly to storage hardware like NetApp, EMC and Exagrid.

Even if you can’t virtualize your custom-built server, you can automate a lot more of the management of it (even if some of that means stopping changes that might otherwise be applied to it by your management systems). The more times you let a human make changes to the system manually, the more chances you have for something to go wrong. Adopting tools like System Center Configuration Manager or PowerShell Desired State Configuration means you’ll have to document the setup of your pet server (which is a good thing) and set policies for how you manage it. You can also use them to reduce the time you spend managing other servers, where you can use a standard configuration and introduce more automation.

Make your snowflake server more reliable by making sure you never run into storage problems. That doesn’t mean the expense of investing in a Storage Area Network (SAN). You can use the pooled storage support built into Windows Server as the Storage Spaces feature to aggregate direct-attached storage – including JBOD enclosures – into tiered, mirrored, resilient storage where it doesn’t matter if a disk fails because the content won’t be lost; a failed disk can be replaced by a hot spare or by capacity on other disks in the pool.

If disk capacity (or the cost of redundant storage) is an issue and your workload is designed for a SAN, you can use a StorSimple appliance to get a local SAN that backs up and fails over to Azure, AWS or OpenStack. That gives you both a way of expanding storage and a disaster recovery system for your server.

If your snowflake server is important enough to keep around, the data in the workload is probably important, too. If you use Delphix’s Data as a Service tools to store a copy of the application data, you can virtualize the data, which lets you track changes and do regression testing. If a change takes the server down, you can give users access to the virtualized data while you revert the system to get it working again. You can also use the virtual data to test changes to your server before you make them to the live system, or to give developers access to the data for their apps without letting them to connect to your fragile server. And when the time comes to migrate away from your snowflake server, having the data virtualized could make it a much simpler project. Delphix works with data sources like SAP, JD Edwards, SharePoint, Oracle (including eBusiness Suite), PeopleSoft and custom apps built with Postgres (and soon MySQL).

It’s important to restrict access to snowflake servers, to ensure that only the people who ought to be using and managing them can actually do so. In particular, make sure that admin accounts for the workloads running of them are separate from general admin accounts. You can do a lot of that with group policy, but it’s also worth looking at privilege management tools. If you’re using Azure AD for single sign-on you can find, restrict and audit admin accounts … but you’re less likely to have that working with these kinds of servers.

Courion’s Access Insight will help you find which accounts are service and admin accounts; Viewfinity and Avecto’s Defendpoint have tools to find admin accounts and restrict privileged access to applications. Or if you use PowerShell and Windows Management Framework 5.0, you can use the Just Enough Admin feature to make sure only specific users on specific systems (even at specific times) can use certain admin commands. Use the JEA Toolkit Helper to set that up.

If you can’t easily load balance or distribute your pet server because it’s so unique, keeping availability high can mean keeping a close eye on the network it’s connected to. Spiceworks Network Monitor is an excellent – and free – network monitoring tool that lets you actively monitor networks and servers, view a dashboard of information (including network utilization, CPU utilization, disk I/O, system memory, active services and processes) and get alerts. It can monitor both Linux and Windows Server, and you can use the range of Spiceworks tools to run inventory reports and keep change management records. If the ads in the free version prove annoying, you can pay to remove them. And Spiceworks users are so enthusiastic, they form a large community where you might find someone else tending a similar server to offer tips and advice.

If you’re trying to diagnose a problem with a pet server, you might have few enough log files to look at by hand. But if you want to dig into more than a few – perhaps to see how what you’re seeing compares to historical data, or to see if the same issues are showing up on other systems – take a look at Splunk. It started as a search engine for log files, Windows event logs, Syslog, SNMP and any other application records you can track down, but it’s added a lot of analytics tools to help you extract information from the mass of raw log data. There are add-ons for a wide range of enterprise applications that help you extract data about how they’re running and analyze it, and there are APIs that you can use to get information from your custom applications as well.