OPS-6220-Ionos-Exporter Extension

[aebyss]

Description

Changes being implemented are big, sorry in advance but I will detail them as I go and link to specific files so it is easier to check and give feedback.
If interested in graphical representation of how functions behave check the Documentation folder with corresponding
Sequence diagrams for s3, postgres.

I will start with ionos-scraper.go

ionos_scraper.go

Here I am doing expanding upon already defined structures and variables in master,
here I defined few help functions which are going to fetch NATGateways

This fetches the NATGateways of given datacenter from IONOS and gives the list back, also we give error back if we cant query the Gateways.

fetch NetworkLoadbalancers

This fetches all of the Networkloadbalancer, same principle as NAT Gateways.

fetch IPBlocks

With fetch IPBlocks we get the all of the IP Addresses in Datacenter, again same principle as with above mentioned.

fetch ApplicationLoadbalancers

Same as the Networkloadblanacers.

Processing of those objects.

Everthing we fetch will also be processed with help of process Helper functions.
Returned values from those process methods are going to be stored in a variable
and then written to the map, which is then used in prometheus collector to iterate throught them and expose that on /metrics Endpoint.
one notable change is here here I have instead os.Exit(1) added a counter variable, which updates a FailedApiRequest metric, and also isntead of exit, I have continue added, it is subject to change I think, but since after 15 minutes of scrape interval this reqeust will be newly generated I tought exiting totally would break all other scraping functions and therefore ionos exporter would be down and we would have a empty result untill it is restarted, there is another ticket which is working on retrying this 3 times and then deciding what to do.

postgres_scraper.go

In postgres metrics we go about same as with ionos-scraper.go where we define structs which encapsulate data which is going to be worked on and then exporter to the prometheus, also we define 2 additional Structs, TelemetryMetric struct which is goign to store all the values, Values is 2D Array which can take any datatype, that is why i have chosen interface and TelemetryResponse is there to capture the structure of telemetry data.

Methods used

PostgresCollectResourcers

This is a main Method, which has the main infinte loop for processing cluster

processCluster

In this function we go through datacenter Items and fetch Owner, and Database names

fetchTelemetryMetrics

Here we return telemetry response, and use it in processCluster function. In processCluster function we create a map in which we store all of those values that are going to be exposed as a metric later.

s3 scraper.go

S3 Scraper uses go aws sdk to query buckets and read objects and their data and then exposes that data to prometheus, data which is scraped is all the methods used on those buckets GET, PUT, POST, HEAD and also their request and response sizes.

Here are defined 2 structs EndpointConfig struct and Metrics struct.
Variables are declared in global scope and are being used for later iterating through Maps and collector will expose that data to /metrics endpoint

createS3ServiceClient

creates a session using AWS (int this example) IONOS access and secret Key which can be generated in Object Storage Console in DCD. Function also gets endpoint and region as argument, it returns new session, this will be used in S3CollectResources to establish 2 connections one for "de" endpoint and another with "eu-central-2' endpoint.

S3CollectResources

here we have infinite loop with sleep wait cycle, we check for already crreated maps with the endpoint if it exists just skip creation and go to establishing new connection, first we are using the Getenv to load IONOS Object Storage keys and check if we have set them or not if not return from function since we can't do anything then. Also we define a endpoints variable which takes a map which is of EndpointConfig datatype and has necessary keys and values of those endpoints.

furthermore we create a semaphore channel which limits how many concurrent processes can be launched,
in the main for loop we establish a client, which takes all of the values of EndpointConfig and then,
this line

basically lists all the buckets, for current owner, that means that all the buckets even those that are not in same endpoints are listed, and this is where AccessDenied error is being thrown if we are not in same endpoint, so basically next step is to iterate over Buckets for which we launch 1 go routine for each bucket, to limit number of concurrent goroutines we use a buffered channel named semaphore.
In a processBucket function it is similar, each object is processed in its own goroutine.

Other functions implemented are the ones for getting getBucketTags(), processObject() and processLine().
In Process Object we are using go AWS SDK functions to get the Object Input where we give its Bucket Name and its Key, on which we then check if we can get it if not then error would be logged and returned from the function, else it makes new buffer in which we are going to stream all the lines of a logfile. When EOF reached break and go to the processLine where we check for all the Methods GET, PUT, POST, HEAD.

Collectors

Every scraper has its own collector, for example ionos_scraper.go has also ionos_collector.go for keeping it readable, since collectors have all the metrics that should be known on compile time, these metrics will then be serverd on /metrics endpoint.
There is nothing complicated about those collectors, they all have basically three methods one constructor to define every descriptor of a metric and returns a pointer to the collector in case of ionos_collector.go it is NewIonosCollector
then there is Describe function which writes all of those descriptors to desc channel and Collect in which we have made iteration over a IonosDatacenters Map and serve those values to prometheus.

Helper.go

Is a file where we defined some structs and functions which are used throughout exporter. GetHeadBucket
makes a request to check if the permissions are ok to check inside that bucket if not it should skip it. So we don't make
lots of get requests that return bigger payload compared to head request.

Config.yaml

The config.yaml is used for all the metrics that are usefull for postgres cluster, I took all the descriptions from IONOS Telemetry website

ConfigMap.yaml

We added here to load the config.yaml charts/ionos-exporter/templates/ConfigMap.yaml

charts/ionos-exporter/templates/deployment.yaml
Here we haev configure how the AWS GO SDK credentials are being handeld, the ACCESS KEY AND SECRET KEY are in 1Pass of monitoring user and is used for S3 scraping.

values.yaml

charts/ionos-exporter/values.yaml

added and commented out token usage. Tought we would need token to authenticate but seems like it works with username and pass still, also variables are added for s3 credentials.

Links to Tickets or other PRs

TIcket OPS-6220
docu confluence
#4

Approval for review

• All points were discussed with the ticket creator, support-team or product owner. The code upholds all quality guidelines from the PR-template.