6. Maintaining

In this project phase, your web site is live and visitors are generating data (e.g., form submissions, web analytics, etc.), editors are producing new content, and you are enhancing the website's features. These are important activities and you need to monitor the website's environment, application, and data in order to maintain peak performance and stay ahead of issues.

Monitoring

Monitoring the website keeps you informed about the health of the site and helps identify what can be optimized and improved.

Environment

  • Monitor resource utilization on the server (e.g., CPU, RAM, HDD).
  • Monitor web site availability with uptime services.
  • Monitor web site responsiveness / performance.
    • Consider checking the performance from different geographic locations.

Application

  • Monitor application errors, warnings, 404 errors, etc.
    • Use event log notifications or create custom report in the reporting module.
  • Monitor application performance with the Kentico Health monitoring performance counters.

Data

  • Keep an eye on web analytics and online marketing data.
    • Monitor all data, like visitors and their activities, page views, conversions, etc.
    • Configure inactive contact deletion rules.
    • Make sure that the scheduled tasks deleting old data are configured and running (e.g., for analytics data and inactive contacts).
  • Consider separating the online marketing data to another database if the size of the data is causing performance issues on the rest of the site.
    • It is more important to keep contact data than web analytics as 3rd party tools (e.g. Google Anayltics) can be used to keep long term analytic data.
    • Keep online marketing data for 6 to 12 months to provide the necessary data for marketing automation and other online marketing features.
  • Monitor the size of the database tables.
  • Define rules to handle outdated content.
  • Monitor content for broken links.

Maintaining

Analyze the data you gather from monitoring to determine what actions to take. Focus on optimizations of content and performance.

Environment maintenance

  • Define backup strategy (SQL, file system, disaster restore / recovery plan).
  • Perform SQL maintenance (shrink transaction log, rebuild indexes).
  • Plan outages and take necessary steps to notify the appropriate people.

Application maintenance

Data maintenance

  • Archive old data.
  • Remove unnecessary objects (e.g., old content versions, unused media files, inactive contacts, old web analytics, staging tasks, unsent emails, old online marketing activities, etc.).
  • Update and optimize content (e.g., update outdated content, fix broken links, perform AB tests, etc.).
  • Deploy new content and changes from staging to production.

For additional information about maintenance in Kentico please refer to the Maintenance in Kentico article.

  • Scenario: Diagnosing performance issue caused by a large amount of content

    A client runs a large news web site with hundreds of thousands of articles in the content tree. Over the time the web site performance rapidly went down.

    When the client checked the SQL server resources, they realized that the usage had increased significantly. Further investigation with SQL Profiler revealed that the issue is caused by having more than 1,000 pages under a single parent page, which is against the best practices.

    To prevent this from happening again, the customer created a custom report in the Reporting application that automatically sends an email notification once per week reporting the status of the content tree. Next, they implemented a new business rule that automatically adjusted and organized the content tree structure to comply with the best practices.

    After this incident, the client also started monitoring resources of all the servers with New Relic, set up the Kentico Health monitoring feature to monitor application performance, and started regularly checking the size of the database tables. Now they will be notified before visitors notice any issues.
  • Scenario: Diagnosing website outages

    A client was experiencing occasional web site outages that they did not notice until their users started complaining how slow and unresponsive the web site was.

    After digging into the event logs, they discovered a pattern where after a content change event, the web site would restart. Whenever an ASP.NET web site is restarted, an extensive compilation might happen. Also, caches have to be primed, which leads to slower page loads after the restart.

    The client decided to precompile their web site. This helped a bit with the compilation slowdown, but did not address the underlying issue. They eventually decided to add third party monitoring software to get more information.

    This lead them to discover that the web site was restarted due to a change in App_Data folder where Kentico stores smart search indexes, web analytics logs, etc. They knew that this should not happen with .NET and eventually it turned out to be a bug in ASP.NET. They found out that this bug was fixed in a Windows update hotfix.

    If the client was monitoring the site and keeping the system up-to-date, they could have avoided these issues entirely.
  • Scenario: Large database causing performance issues

    A client was utilizing nearly all of Kentico's features including online marketing tracking, web analytics, version history for objects and pages, workflow, and many others.

    Since they did not have any data monitoring or a maintenance strategy, the size of the database was growing rapidly (100+GB) and started slowing down the system. Eventually, the added load exceeded the capabilities of their environment and the site went down.

    In this case, The web site was a large web site with hundreds of thousands of pages in the content tree, each with an average of 100 versions, and over 100 million contacts, activities, and web analytics entries.

    Once they got hte data cleaned up, they implemented a plan to monitor the database size and manage data growth using built-in features like deleting inactive contacts, limiting the version history, and clearing web analytics data on a regular bases.

References