Submind YouTube summaries
Thumbnail for HA Looked Fine. Then the Outage Hit. | Cassius Rhue, SIOS Technology

HA Looked Fine. Then the Outage Hit. | Cassius Rhue, SIOS Technology

Watch on YouTube

Video summary

The video highlights the critical importance of routine High Availability (HA) testing by presenting real-world examples where such tests uncovered hidden configuration errors before they could cause significant downtime. In one recent instance, a partner's system was found to have a protected service misconfigured to start automatically outside of the HA solution's control. This setting would have served stale data or blocked access entirely during a failover, but early detection allowed the team to correct the configuration and ensure the system always pointed to the latest data. Similarly, another case revealed that client applications were hardcoded to connect to the physical server IP address rather than the virtual IP used during failovers; without testing, this would have prevented clients from connecting once services moved to the backup site. A third example involved a primary system upgrade where a configuration file and necessary packages were updated on the primary node but forgotten on the backup, an issue that routine validation would have caught and resolved before a disaster occurred. Conversely, the transcript details a stark scenario where the absence of regular testing led to a major business-critical outage despite the HA environment appearing healthy. A client declined external validation services and skipped internal testing during a maintenance window, while also lacking a dedicated test environment for updates. Consequently, when a power outage forced a failover, the backup server failed because it was missing updated libraries and packages that existed only on the primary system. Furthermore, configuration parameters stored locally on the primary node had not been replicated to the secondary server. This combination of missing dependencies and unsynced settings turned what should have been a seamless transition into a complex emergency requiring multiple teams to troubleshoot database errors and application startup failures simultaneously. The conclusion drawn from these contrasting experiences is that routine testing acts as a vital preventative measure that saves businesses from chaos, panic, and financial loss during actual disasters. When an outage occurs without prior validation, organizations are forced into reactive troubleshooting modes where they scramble to identify root causes, coordinate with multiple vendors, and often rely on on-call engineers who may not be fully familiar with the system's nuances. This delay allows executives to watch their business lose money by the second while teams work through issues that could have been resolved in minutes during a scheduled test window. Ultimately, the video emphasizes that proactive validation ensures seamless failovers, maintains data integrity, and prevents the costly confusion that arises when critical infrastructure fails without warning.
Read the full video transcript
Is it possible for you to share a real world example where routine HA testing uncovered a hidden issue that would have only surfaced during an actual outage and would have actually caused downtime and chaos? >> Yeah, we actually have um I have two um in fact one is very recent as recent as earlier this week. Um so in the first one we were actually working with a partner who was setting up a system for an end customer and they engaged with us to do some testing to make sure that after their installation and updates that the system was working correctly. And our very first validation test uh we discovered that one of the protected services was uh misconfigured. In this particular uh case the service was set to what's called automatic start. So it actually would start outside of the HA solution and when that occurs uh the the the data that it would serve up would be stale data um because it was not you know a part of the HA monitoring and recovery. By detecting that early, we were able to go in change the configuration so that it was would not start outside of the HA solutions control and make sure that it was always going to be pointing to the the latest set of data. You know, if that had been a found during a realworld scenario, just think of how long it may have been before the end customer discovered that they were receiving stale data. um or in in some examples uh related to this particular environment, the client application can start in in a way in which it no longer allows anyone else to access the data. And so finding that prevents an outage. Um it prevents an error. It prevents data integrity uh issues because um you've detected it early. you've corrected the configuration and then you went on to further validate that things are working properly with the HA solution you know um a second one related to that uh to a client realizing that they had configured all of their client software so our end customer had an HA solution in place uh using our SCIOS lifekeeper high availability solution they were protecting several databases and applications uh their clients however ever were configured originally to point to the physical server IP address. And so once the applications, services, and databases failed over to the backup, those clients could not connect because they had been hardcoded. Testing and validating in the the U customer's environment helped us find that early, make that a change and adjustment and then get things corrected and retested in a way in which the failover and client reconnectivity became seamless. And that's something you don't want to find during an actual outage because of course uh when there's an outage you you're having to troubleshoot. You're having to do a lot more root cause analysis. You're opening tickets with multiple vendors. You're trying to coordinate understanding and and what are the symptoms and signs and problems and one thing that often gets overlooked if the applications or databases are up and running uh then then you're having to go look and see okay what happened with clients and that that and that panic and chaos and that moment where you're scrambling is something that can be missed. So testing in real world scenarios helped us find that before it became an issue in in an actual disaster. And then one recently we had a client that did an upgrade of their primary system but forgot to actually do the upgrade on the backup system. And they called us in because they wanted to do some testing. And what we found is that there was a configuration file that needed to be um updated on both servers. And that was something that we needed to make sure that they were able to run through and that allowed us to to find the issue, update the configuration file and also update their runbook. So now going forward they know uh the procedures and steps they have to do for updates of the operating system, updates of their critical database, updates of their client application and they know how to test it in a proper manner and where to validate that things were done uh correctly and so that saves them a lot of time. Of course, if you're in an actual disaster, uh you're you're looking at, okay, well, why didn't this work? And you're starting with troubleshooting steps. And often these are the little uh nuances that you might miss while there are 21 executives wondering why their business is losing money by the second. Now, let's flip this coin. Can you walk us through a scenario where an HA environment look perfectly healthy but the absence of routine testing led to real problems when a failover event actually happened? That's a great question and it's a very vivid example in my mind because of uh how how many um different organizations were involved in this particular incident and how long um it took to resolve and remediate the problem and the fact that it was a business critical system in a live production down event. Uh so we did have a client who um set up their system with our software and they went through some uh what they considered to be routine maintenance. It was actually turned out to be more than routine maintenance. Uh but for time constraint pressures or business pressures or stakeholder involved pressure uh this particular client opted you know first uh to decline services that we offer for validation and and uh testing and then second they also declined to do that testing themselves to make sure during their maintenance window that they um that the failover would work or that the applications would run on the backup. A third thing that we've kind of touched on earlier, they also did not have a test environment where they could run these updates uh prior to rolling them out to production. And so all of the updates were made on their production system. Things appeared to be healthy on the production node. Um they had an outage, a power outage in their data center. uh things uh attempted to fail over to the backup server and that's when they discovered that there were during the upgrade of the primary there were some configuration changes that were made. Uh there were also some libraries and packages that had been updated to a newer version that did not exist on their uh target system or their secondary. And so when it came time to start the the database, uh the database failed with some errors related to those updated packages that were missing. Uh the the application when they were able to troubleshoot the database and get it to a running and working state. uh the actual application failed because there were some configuration parameters that had been made on the primary um and stored locally in a in a config file had not been replicated or synced to the backup server. And so in a scramble you you have a a team member uh from their organization from the application team trying to troubleshoot what's going on, why the application won't start. Um, a lot of times you you have in an emergency like that, you may not have the architect of the application available. And so you're dealing with an on call engineer from the app team, uh, an on call engineer from the database team. Uh, you may be dealing with folks that are not as familiar with the system. And in that emergency situation, uh, this particular customer was trying to figure all of these things out at one time. you know uh ultimately their resolution was to update the packages on that target system, update the configuration file and update the database so that it would all start and run. That outage um instead of it being a simple failover, you know, power failed in the primary data center, everything starts up seamlessly on the backup without clients noticing it. instead that turned into a major outage event and cost them uh significantly. Whereas a routine test uh would have identified all of these issues and would have saved the business a lot of money.