Graylog, headquartered in Houston, offers their eponymous platform for centralized log management that helps users find meaning in data faster so as to take action immediately. Graylog is available via Enterprise and Cloud plans, but also has a Small Business Plan, and an Open (free) plan with limited features.
Logstash can be compared to other ETL frameworks or tools, but it is also complementary to several, for example, Kafka. I would not only suggest using Logstash when the rest of the ELK stack is available, but also for a self-hosted event collection pipeline for various …
For small companies, Graylog is the best solution possible. It's easy to configure and "just works." Above everything else, it's free. The only thing I hold against it is the fact that it's Linux-based. [This] makes sense because Elasticsearch is Linux-based. But Linux adds a layer of complexity that we don't need for something basic as a logging server. I'm pretty sure that we would have had a logging server years earlier if I had to convince quite a few decision-making people to go ahead with it anyway.
Perfect for projects where Elasticsearch makes sense: if you decide to employ ES in a project, then you will almost inevitably use LogStash, and you should anyways. Such projects would include: 1. Data Science (reading, recording or measure web-based Analytics, Metrics) 2. Web Scraping (which was one of our earlier projects involving LogStash) 3. Syslog-ng Management: While I did point out that it can be a bit of an electric boo-ga-loo in finding an errant configuration item, it is still worth it to implement Syslog-ng management via LogStash: being able to fine-tune your log messages and then pipe them to other sources, depending on the data being read in, is incredibly powerful, and I would say is exemplar of what modern Computer Science looks like: Less Specialization in mathematics, and more specialization in storing and recording data (i.e. Less Engineering, and more Design).
Graylog does a great job of its core function: log aggregation, retention, and searching.
Graylog has a very flexible configuration. The backend for storage is Elasticsearch and MongoDB is used to store the configuration. You have to option to make your configuration as simple as possible by storing everything on one box, or you can scale everything out horizontally by using a cluster of Elasticsearch nodes and MongoDB servers with several Graylog servers pointed to all the necessary nodes.
Graylog does a good job of abstracting away a fair portion of Elasticsearch index management (sharding, creation, deletion, rotation, etc).
Logstash design is definitely perfect for the use case of ELK. Logstash has "drivers" using which it can inject from virtually any source. This takes the headache from source to implement those "drivers" to store data to ES.
Logstash is fast, very fast. As per my observance, you don't need more than 1 or 2 servers for even big size projects.
Data in different shape, size, and formats? No worries, Logstash can handle it. It lets you write simple rules to programmatically take decisions real-time on data.
You can change your data on the fly! This is the CORE power of Logstash. The concept is similar to Kafka streams, the difference being the source and destination are application and ES respectively.
Since it's a Java product, JVM tuning must be done for handling high-load.
The persistent queue feature is nice, but I feel like most companies would want to use Kafka as a general storage location for persistent messages for all consumers to use. Using some pipeline of "Kafka input -> filter plugins -> Kafka output" seems like a good solution for data enrichment without needing to maintain a custom Kafka consumer to accomplish a similar feature.
I would like to see more documentation around creating a distributed Logstash cluster because I imagine for high ingestion use cases, that would be necessary.
Community support does not give simple straightforward answers; simply search up Graylog Issues and look at some of the responses on the forums. The documentation is your only hope if you are on the free version, as you can NOT purchase only support. The few times I have worked with Graylog Enterprise support they were great though.
In terms of log aggregation, the free product fully stacks up with the competitors listed. Full control over the data ingests for flexible configuration. Graylog even better on that front than AlienVault USM because you cannot configure the variable mapping. We haven't used the threat exchange stuff or correlation. But with regex searches, we have created function dashboards that show threat theater pictures of our network based on logs from our firewall.
MongoDB and Azure SQL Database are just that: Databases, and they allow you to pipe data into a database, which means that alot of the log filtering becomes a simple exercise of querying information from a DBMS. However, LogStash was chosen for it's ease of integration into our choice of using ELK Elasticsearch is an obvious inclusion: Using Logstash with it's native DevOps stack its really rational
Graylog is just less expensive than some other options which meant it fit into our budget otherwise we might not be able to justify a higher cost.
Being able to track issues that we normally couldn't track using other tools is a bonus to help us know of any issues we have and can fix before an outage or failure that could potentially cost money.
We have had to spend more time than I would like to understand and customize Graylog which has taken time away from other tasks and projects.
Positive: Learning curve was relatively easy for our team. We were up and running within a sprint.
Positive: Managing Logstash has generally been easy. We configure it, and usually, don't have to worry about misbehavior.
Negative: Updating/Rehydrating Logstash servers have been little challenging. We sometimes even loose data while Logstash is down. It requires more in-depth research and experiments to figure the fine-grained details.
Negative: This is now one more application/skill/server to manage. Like any other servers, it requires proper grooming or else you will get in trouble. This is also a single point of failure which can have the ability to make other servers useless if it is not running.
