Time is relative? Yes, it is. When you need to write Audit Data and Logs, you don’t want to have a relative time on different systems.
There are different protocalls for time synchronization like NTP that can help us to synchronize the time on machines. Unfortunately, based on firewall configuration or the type of access we have on that machines we might not be able to use NTP protocol.
Even if we are using NTP, depending on server configuration we can have a deviation of 1-2s between machines. A deviation of a few seconds might not be too much until you write logs and you try them to order them based on time. In this situation, reading the logs is not so simple – the logs order will not match with the execution order of the actions that were logged. Trying to understand what happen inside the system wil be hard.
If this is not enough, we might have cases when our application is running in different environments that are using different time servers. For example, we can have a group of systems that are running on Azure machines, another group of systems that are running on-premises and another set of machines from AWS. The time will not match perfectly. Each server will have a small time difference.
Usualy all NTP servers are sync between each others, but still, there can be a small time difference. The biggest problem is with on-premises systems, where the deviation can be 1s or more.
What we could do in this situation? Our mission is to be able to order logs from a timeline perspective, in the correct order.
From the time perspective, logs that were generated by the same system can be ordered correctly using the time information.
Another important information that we have is that we know the flow of data and logic inside the system. This means that we know that a request is first processed by system A, after it will be processed by system B and so on.
Even if we have cycles in the flow between different system, we can consider that when the current requests arrives in a system where it was already, we can view it as a new system.
Using this approach, we define ‘gates’ inside our system that can be tracked and numbered. If each gate has a different number or id, we are able to order them based on this information. In this way we can put all our logs in only one repository, order them based on the system (gate) number and each group of logs can be ordered based on the time.
Different tracking ID of the request will help us to identify from the logs repository the request that we want to analyze.
In the end we will have the right order of our logs and we’ll be able to read them like a story. For complex scenarios, we might need to work a little more, but the main idea will remain the same.
There are different protocalls for time synchronization like NTP that can help us to synchronize the time on machines. Unfortunately, based on firewall configuration or the type of access we have on that machines we might not be able to use NTP protocol.
Even if we are using NTP, depending on server configuration we can have a deviation of 1-2s between machines. A deviation of a few seconds might not be too much until you write logs and you try them to order them based on time. In this situation, reading the logs is not so simple – the logs order will not match with the execution order of the actions that were logged. Trying to understand what happen inside the system wil be hard.
If this is not enough, we might have cases when our application is running in different environments that are using different time servers. For example, we can have a group of systems that are running on Azure machines, another group of systems that are running on-premises and another set of machines from AWS. The time will not match perfectly. Each server will have a small time difference.
Usualy all NTP servers are sync between each others, but still, there can be a small time difference. The biggest problem is with on-premises systems, where the deviation can be 1s or more.
What we could do in this situation? Our mission is to be able to order logs from a timeline perspective, in the correct order.
From the time perspective, logs that were generated by the same system can be ordered correctly using the time information.
Another important information that we have is that we know the flow of data and logic inside the system. This means that we know that a request is first processed by system A, after it will be processed by system B and so on.
Even if we have cycles in the flow between different system, we can consider that when the current requests arrives in a system where it was already, we can view it as a new system.
Using this approach, we define ‘gates’ inside our system that can be tracked and numbered. If each gate has a different number or id, we are able to order them based on this information. In this way we can put all our logs in only one repository, order them based on the system (gate) number and each group of logs can be ordered based on the time.
Different tracking ID of the request will help us to identify from the logs repository the request that we want to analyze.
In the end we will have the right order of our logs and we’ll be able to read them like a story. For complex scenarios, we might need to work a little more, but the main idea will remain the same.
Comments
Post a Comment