Monitoring After Deployment

  • I'm more of a developer role, but there is only me, so I have to act as the DBA too.

    I normally measure the impact of my queries when developing them before and after deployment. I only use tools provided by Microsoft outside of SQL Prompt, which is just used to make development a bit more efficient.

    I think the one area that I did want to monitor long term though was my ETL Systems. I created a few automated jobs that send daily reports on the health of SQL Server and the OS. This includes daily stats on memory, disk usage, record counts, times of reading and writing data to and from the various services.

    I also have a pretty detailed log system, which I actually adapted from this site, which logs and sends updates between each stage of my ETL process.

    On the Python level, where I deal with my API's, I also have pretty extensive logging. That way I can pinpoint issues down to the API level when SQL is not telling me the full story.

    I would suspect that any deployment therefore after, would likely show through those systems in one way or another if the internal monitoring tools do not. This includes delays, bottlenecks and so forth that impact the entire environment.

  • I developed windows services to monitor a fragile vendor web application that sent out notifications and alerts, and a Linux based hardware appliance that processed HL7 messages that was also prone to crashes.  The services ran on a specified interval and would make sure that the web server returned a login page (looking for the user name and password fields) and also verify that it could connect to the databases. The services would notify the appropriate team, either server hosting or the DBAs, of errors encountered. The service that monitored the notification system would check for data corruption in a column since code executed this SQL statement: select MAX(Convert(Int,EventID)) from Envents; a non-numeric value in the EventID column would cause the system to crash. For unknown reasons, date/time columns in the vendor's databases were defined as character and not datetime.

    Since then, I've used the services shell to create other monitoring services for systems that I developed or support.

Viewing 2 posts - 16 through 16 (of 16 total)

You must be logged in to reply to this topic. Login to reply