Welcome!

Linux Containers Authors: Liz McMillan, Yeshim Deniz, Elizabeth White, Pat Romanski, Zakia Bouachraoui

Related Topics: @DevOpsSummit, Java IoT, Microservices Expo, Linux Containers, @CloudExpo, @DXWorldExpo

@DevOpsSummit: Blog Post

Software Quality Metrics for Your Continuous Delivery Pipeline | Part 3

It is safe to say that insightful logging and performance are two opposite goals

Let me ask you a question: would you say that you have implemented logging correctly for your application? Correct in the sense that it will provide you with all the insights you require to keep your business going once your users are struck by errors? And in a way that does not adversely impact your application performance? Honestly, I bet you have not. Today I will explain why you should turn off logging completely in production because of its limitations:

  • Relies on Developers
  • Lacks Context
  • Impacts Performance

Intrigued? Bear with me and I will show you how you can still establish and maintain a healthy and useful logging strategy for your deployment pipeline, from development to production, guided by metrics.

What Logging Can Do for You
Developers, including myself, often write log messages because they are lazy. Why should I set a breakpoint and fire up a debugger if it is so much more convenient to dump something to my console via a simple println()? This simple yet effective mechanism also works on headless machines where no IDE is installed, such as staging or production environments:

System.out.println("Been here, done that.");

Careful coders would use a logger to prevent random debug messages from appearing in production logs and additionally use guard statements to prevent unnecessary parameter construction:

if (logger.isDebugEnabled()) {
logger.debug("Entry number: " + i + " is " + String.valueOf(entry[i]));
}

Anyways, the point about logging is that that traces of log messages allow developers to better understand what their program is doing in execution. Does my program take this branch or that branch? Which statements were executed before that exception was thrown? I have done this at least a million of times, and most likely so have you:

if (condition) {
logger.debug("7: yeah!")
} else {
logger.debug("8: DAMN!!!")
}

Test Automation Engineers, usually developers by trade, equally use logging to better understand how the code under test complies with their test scenarios:

class AgentSpec extends spock.lang.Specification {

def "Agent.compute()"() {
def agent = AgentPool.getAgent()

when:
def result = agent.compute(TestFixtures.getData())

then:
logger.debug("result: "  + result);
result == expected
}
}

Logging is, undoubtedly, a helpful tool during development and I would argue that developers should use it as freely as possible if it helps them to understand and troubleshoot their code.

In production, application logging is useful for tracking certain events, such as the occurrence of a particular exception, but it usually fails to deliver what it is so often mistakenly used for: as a mechanism for analyzing application failures in production. Why?

Because approaches to achieving this goal with logging are naturally brittle: their usefulness depends heavily on developers, messages are without context, and if not designed carefully, logging may severely slow down your application.

Secretly, what you are really hoping to get from your application logs, in the one or the other form, is something like this:

A logging strategy that delivers out-of-the-box using dynaTrace: user context, all relevant data in place, zero config

The Limits of Logging
Logging Relies on Developers
Let's face it: logging is, inherently, a developer-centric mechanism. The usefulness of your application logs stands and falls with your developers. A best practice for logging in production says: "don't log too much" (see Optimal Logging @ Google testing blog). This sounds sensible, but what does this actually mean? If we recall the basic motivation behind logging in production, we could equally rephrase this as "log just enough information you need to know about a failure that enables you to take adequate actions". So, what would it take your developers to provide such actionable insights? Developers would need to correctly anticipate where in the code errors would occur in production. They would also need to collect any relevant bits of information along an execution path that bear these insights and, last but not least, present them in a meaningful way so that others can understand, too. Developers are, no doubt, a critical factor to the practicality of your application logs.

Logging Lacks Context
Logging during development is so helpful because developers and testers usually examine smaller, co-located units of code that are executed in a single thread. It is fairly easy to maintain an overview under such simulated conditions, such as a test scenario:

13:49:59 INFO com.company.product.users.UserManager - Registered user ‘foo'.
13:49:59 INFO com.company.product.users.UserManager - User ‘foo' has logged in.
13:49:59 INFO com.company.product.users.UserManager - User ‘foo' has logged out.

But how can you reliably identify an entire failing user transaction in a real-life scenario, that is, in a heavily multi-threaded environment with multiple tiers that serve piles of distributed log files? I say, hardly at all. Sure, you can go mine for certain isolated events, but you cannot easily extract causal relationships from an incoherent, distributed set of log messages:

13:49:59 INFO com.company.product.users.UserManager - User ‘foo' has logged in.
13:49:59 INFO com.company.product.users.UserManager - User ‘bar' has logged in.
...
13:49:60 SEVERE org.hibernate.exception.JDBCConnectionException: could not execute query
at org.hibernate.exception.SQLStateConverter.convert(SQLStateConverter.java:99)
...

After all, the ability to identify such contexts is key to deciding why a particular user action failed.

Logging Impacts Performance
What is a thorough logging strategy worth if your users cannot use your application because it is terribly slow? In case you did not know, logging, especially during peak load times, may severely slow down your application. Let's have a quick look at some of the reasons:

Writing log messages from the application's memory to persistent storage, usually to the file system, demands substantial I/O (see Top Performance Mistakes when moving from Test to Production: Excessive Logging). Traditional logger implementations wrote files by issuing synchronous I/O requests, which put the calling thread into a wait state until the log message was fully written to disk.

In some cases, the logger itself may cause a decent bottle-neck: in the Log4j library (up to version 1.2), every single log activity results in a call to an internal template method Appender.doAppend() that is synchronized for thread-safety (see Multithreading issues - doAppend is synchronised?). The practical implication of this is that threads, which log to the same Appender, for example a FileAppender, must queue up with any other threads writing logs. Consequently, the application spends valuable time waiting in synchronization instead of doing whatever the app was actually designed to do. This will hurt performance, especially in heavily multi-thread environments like web application servers.

These performance effects can be vastly amplified when exception logging comes into play: exception data, such as error message, stack trace and any other piggy-backed exceptions ("initial cause exceptions") greatly increase the amount of data that needs to be logged. Additionally, once a system is in a faulty state, the same exceptions tend to appear over and over again, further hurting application performance. We had once monitored a 30% drawdown on CPU resources due to more than 180,000 exceptions being thrown in only 5 minutes on one of our application servers (see Performance Impact of Exceptions: Why Ops, Test and Dev need to care). If we had written these exceptions to the file system, they would have trashed I/O, filled up our disk space in no time and had considerably increased our response times.

Subsequently, it is safe to say that insightful logging and performance are two opposite goals: if you want the one, then you have to make a compromise on the other.

For more logging tips click here for the full article.

More Stories By Martin Etmajer

Leveraging his outstanding technical skills as a lead software engineer, Martin Etmajer has been a key contributor to a number of large-scale systems across a range of industries. He is as passionate about great software as he is about applying Lean Startup principles to the development of products that customers love.

Martin is a life-long learner who frequently speaks at international conferences and meet-ups. When not spending time with family, he enjoys swimming and Yoga. He holds a master's degree in Computer Engineering from the Vienna University of Technology, Austria, with a focus on dependable distributed real-time systems.

Comments (0)

Share your thoughts on this story.

Add your comment
You must be signed in to add a comment. Sign-in | Register

In accordance with our Comment Policy, we encourage comments that are on topic, relevant and to-the-point. We will remove comments that include profanity, personal attacks, racial slurs, threats of violence, or other inappropriate material that violates our Terms and Conditions, and will block users who make repeated violations. We ask all readers to expect diversity of opinion and to treat one another with dignity and respect.


IoT & Smart Cities Stories
Contextual Analytics of various threat data provides a deeper understanding of a given threat and enables identification of unknown threat vectors. In his session at @ThingsExpo, David Dufour, Head of Security Architecture, IoT, Webroot, Inc., discussed how through the use of Big Data analytics and deep data correlation across different threat types, it is possible to gain a better understanding of where, how and to what level of danger a malicious actor poses to an organization, and to determin...
The hierarchical architecture that distributes "compute" within the network specially at the edge can enable new services by harnessing emerging technologies. But Edge-Compute comes at increased cost that needs to be managed and potentially augmented by creative architecture solutions as there will always a catching-up with the capacity demands. Processing power in smartphones has enhanced YoY and there is increasingly spare compute capacity that can be potentially pooled. Uber has successfully ...
The deluge of IoT sensor data collected from connected devices and the powerful AI required to make that data actionable are giving rise to a hybrid ecosystem in which cloud, on-prem and edge processes become interweaved. Attendees will learn how emerging composable infrastructure solutions deliver the adaptive architecture needed to manage this new data reality. Machine learning algorithms can better anticipate data storms and automate resources to support surges, including fully scalable GPU-c...
We are seeing a major migration of enterprises applications to the cloud. As cloud and business use of real time applications accelerate, legacy networks are no longer able to architecturally support cloud adoption and deliver the performance and security required by highly distributed enterprises. These outdated solutions have become more costly and complicated to implement, install, manage, and maintain.SD-WAN offers unlimited capabilities for accessing the benefits of the cloud and Internet. ...
Dion Hinchcliffe is an internationally recognized digital expert, bestselling book author, frequent keynote speaker, analyst, futurist, and transformation expert based in Washington, DC. He is currently Chief Strategy Officer at the industry-leading digital strategy and online community solutions firm, 7Summits.
As IoT continues to increase momentum, so does the associated risk. Secure Device Lifecycle Management (DLM) is ranked as one of the most important technology areas of IoT. Driving this trend is the realization that secure support for IoT devices provides companies the ability to deliver high-quality, reliable, secure offerings faster, create new revenue streams, and reduce support costs, all while building a competitive advantage in their markets. In this session, we will use customer use cases...
Machine learning has taken residence at our cities' cores and now we can finally have "smart cities." Cities are a collection of buildings made to provide the structure and safety necessary for people to function, create and survive. Buildings are a pool of ever-changing performance data from large automated systems such as heating and cooling to the people that live and work within them. Through machine learning, buildings can optimize performance, reduce costs, and improve occupant comfort by ...
René Bostic is the Technical VP of the IBM Cloud Unit in North America. Enjoying her career with IBM during the modern millennial technological era, she is an expert in cloud computing, DevOps and emerging cloud technologies such as Blockchain. Her strengths and core competencies include a proven record of accomplishments in consensus building at all levels to assess, plan, and implement enterprise and cloud computing solutions. René is a member of the Society of Women Engineers (SWE) and a m...
With 10 simultaneous tracks, keynotes, general sessions and targeted breakout classes, @CloudEXPO and DXWorldEXPO are two of the most important technology events of the year. Since its launch over eight years ago, @CloudEXPO and DXWorldEXPO have presented a rock star faculty as well as showcased hundreds of sponsors and exhibitors! In this blog post, we provide 7 tips on how, as part of our world-class faculty, you can deliver one of the most popular sessions at our events. But before reading...
Poor data quality and analytics drive down business value. In fact, Gartner estimated that the average financial impact of poor data quality on organizations is $9.7 million per year. But bad data is much more than a cost center. By eroding trust in information, analytics and the business decisions based on these, it is a serious impediment to digital transformation.