#sre - November 2019

November 2019
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)about 7 years ago
You might consider using kinesis
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)about 7 years ago
Stream the logs first there
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)about 7 years ago
Then from there stream them to cloud watch and s3
Sergey
Sergeyabout 7 years ago(edited)
Erik, thanks for your answer. So, we already use Kinesis for streaming logs to s3 but Kinesis does not have the ability to stream logs from s3
I think we should use some Lambda function for this purpose, the same way as DataDog uses for getting logs from customer's s3 buckets
But anyway thanks a lot
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)about 7 years ago
No, you don’t stream from s3
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)about 7 years ago(edited)
But you can output to both s3 and cloud watch logs at the same time
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)about 7 years ago
Which is I think what you wanted to achieve
Sergey
Sergeyabout 7 years ago
Yes, that is I want, but looks like kinesis can not stream to CloudWatch
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)about 7 years ago
Hrmm
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)about 7 years ago
Look into AWS glue?
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)about 7 years ago
@Andriy Knysh (Cloud Posse) might know more
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)about 7 years ago
He worked on something similar
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)about 7 years ago
I want to implement something similar. Stream to s3 and elastic search
Sergey
Sergeyabout 7 years ago
Yeah, I saw this manual
https://aws.amazon.com/ru/blogs/security/enabling-serverless-security-analytics-using-aws-waf-full-logs/
But it brings several additional tools and looks like very overloaded
Sergey
Sergeyabout 7 years ago
I have found this lambda function https://github.com/miztiik/serverless-s3-to-elasticsearch-ingester
So, pyton code very similar to the lambda function that datadog uses)
I am going to update that codebase and add additional function - streaming to CloudWatch
Andriy Knysh (Cloud Posse)
Andriy Knysh (Cloud Posse)about 7 years ago
why do you need to move it from S3 to CloudWatch? What’s the use case?
Sergey
Sergeyabout 7 years ago
Hi, Andriy. Nice to see you
We are using AWS ALB + WAF. They can streams their logs only to s3 bucket. Also we are using CloudWatch as one point for monitoring and logging our applications and infrastructure components. So, we need some tools that can get logs from s3 and send them to CloudWatch
Sergey
Sergeyabout 7 years ago
I want to replace ELK stack that we use for log collecting in favour of Fluentd + CloudWatch. Now I can collect logs from customer application and kubernetes cluster to CW, also we use CW Container Insights as Kubernetes monitoring tools
Andriy Knysh (Cloud Posse)
Andriy Knysh (Cloud Posse)about 7 years ago
hey Sergey, yes looks like lambda is the only way to go to move logs from S3 to CW
Gareth
Garethabout 7 years ago(edited)
Good morning All,
I'm looking for some advice around monitoring and wondered if somebody had the time to give me their opinions.
I work for a company that's got multiple AWS accounts, One company account (master) and then one per client which contain many VPC's normal separated based on DEV, Test, Pre-Live and Prod per product we design. Primarily web sites or applications.
We've historically used Cloudwatch and created the dashboards as an when we need something above and beyond the default EC2 monitoring but this is become costly and doesn't really give us a centralised approach or the flexibility to add something that is beyond the basics.
I'm looking to bring us in to a more centralised way of working and have been reviewing Prometheus + Grafana and maybe Thanos to get a more historical view of data. I've seen the SweetOps articles on Prometheus and read over the kubernetes example but I'm left confused as to what is the best way to architect a Prometheus platform. Most examples I've seen look to deploy Prometheus on kubernetes, but I would have thought it would be better off as an EC2 node. There also appears to be a split in the thinking that its would be best to deliver a Prometheus pair of server to each AWS account rather than trying to bridge from say the Companies Master account. Appreciate this is a wide and very generalized question but at this time any comments would be of help to me.
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)about 7 years ago
@Gareth those docs are a bit out of date
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)about 7 years ago
Sorry
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)about 7 years ago
Join us this coming Wednesday for #office-hours and we can discuss
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)about 7 years ago
Ennio
Ennioalmost 7 years ago(edited)
Hi all, anyone set up webpage client side metrics getting to Prometheus? I wanted to build a poc and I've found this https://www.weave.works/blog/browser-metrics-into-prometheus/.
Does anyone have a different solution that would like to share, just to get some different options
Mitch
Mitchalmost 7 years ago
Hey all, I'm looking to use https://github.com/cloudposse/prometheus-to-cloudwatch. Trying to figure out what my cloudwatch_namespace is. Any advice on figuring it out?
Mitch
Mitchalmost 7 years ago
My service is running in ECS with a FARGATE deployment
Erik Osterman (Cloud Posse)
Erik Osterman (Cloud Posse)almost 7 years ago
@Andriy Knysh (Cloud Posse) ☝️
Andriy Knysh (Cloud Posse)
Andriy Knysh (Cloud Posse)almost 7 years ago
Andriy Knysh (Cloud Posse)
Andriy Knysh (Cloud Posse)almost 7 years ago