Google Cloud's operations suite (formerly Stackdriver) Reviews

The features that I have found most valuable are its graphs - if I need any statistics, in Kubernetes or at a VPN level, I can quickly get the reports.

As far as what can be improved in Stackdriver, on the application side, if I want to track any round-trip or breakdowns of my response times, I'm not able to get it. My request goes through various levels of the Google Cloud Platform (GCP) and comes back to my client machine. Suppose that my request has taken 10 seconds overall, so if I want to break it down, to see where the delay is happening within my architecture, I am not able to find that out using Stackdriver. It can give me the utilization reports, and other network traffic information, but it is not able to provide me answers in detail at a request level. That is why I have to rely on some other APM tools.

It would be better if they could provide the round-trip of the response time at a request level because on an overall basis it gives the traffic, transactions per second, etc. if any drops happen, what could be the reason for the drops? Suppose I have a Kong, Kubes,NAT and VPN. With these kinds of things, If any drop happened, it will show me that the drop happened, but it will not give me the reason why it got dropped.

I have been using Google Stackdriver for around six months.

Google Stackdriver is good because when you are using the GCP, there is no need to rely on any other monitoring tools, it gives the majority of the reports. But when you're stuck with any issue, and the team needs to do fine-tuning, it is a little difficult to analyze, especially while performing the load or stress tests. If everything is good, then you can get the reports. However, when something happens, they have to improve reporting.

On a scale of one to ten, I can give Google Stackdriver an eight. 

As I said in our previous discussion, we are finding it difficult to develop the breakdown of the round-trip at a request level, especially pertaining to the response time. That's why if any drops happen in the network layer, it would be good if it could find out and then come up with some reasons, various reasons, potential reasons, why that drop happened and where the drop happened. Also, at the VPN level, there is a replay issue happening. Suppose the request is coming multiple times, we'll call it a replay issue. So to monitor the replay issue, if they would provide any specific functionality it would be good.

We have some projects on the Google Cloud Platform and we are looking into how we can use Stackdriver to monitor our workloads.

The most valuable feature is the multi-cloud integration, where there is support for both GCP and AWS. 

The APM functionality needs to be improved.

It is difficult to estimate in advance how much something is going to cost. Often, you don't know how much it costs until you are finished.

We have been working with Stackdriver for less than one year.

My impression is that this is a stable solution because we have only had a few minor problems.

Stackdriver is very scalable.

The technical support is good and I found them to be very helpful.

The implementation involves setting up the correct metrics, which is something that needs to be properly researched and tested.

The cost of using Stackdriver depends on usage.

My advice to anybody who is implementing Google Stackdriver is to explore what metrics they want to use.

I would rate this solution an eight out of ten.