jmal98 / sqs-exporter Goto Github PK

It seems that the exported metrics are not compatible with older versions of prometheus. We unfortunately are still using version 1.1.2 and it complains about the following error:

text format parsing error in line 123: second HELP line for metric name "sqs_approximatenumberofmessagesdelayed"

I suspect, that it is due to duplicated TYPE and HELP hints.

Do you think it is possible to have them only written once?

The exported metrics data looks like this:

# HELP sqs_approximatenumberofmessagesdelayed The approximate number of messages that are waiting to be added to the queue.
# TYPE sqs_approximatenumberofmessagesdelayed gauge
sqs_approximatenumberofmessagesdelayed{queue="queue-1",} 0.0
# HELP sqs_approximatenumberofmessages The approximate number of visible messages in a queue.
# TYPE sqs_approximatenumberofmessages gauge
sqs_approximatenumberofmessages{queue="queue-1",} 0.0
# HELP sqs_approximatenumberofmessagesnotvisible The approximate number of messages that have not timed-out and aren't deleted.
# TYPE sqs_approximatenumberofmessagesnotvisible gauge
sqs_approximatenumberofmessagesnotvisible{queue="queue-1",} 0.0
# HELP sqs_approximatenumberofmessagesdelayed The approximate number of messages that are waiting to be added to the queue.
# TYPE sqs_approximatenumberofmessagesdelayed gauge
sqs_approximatenumberofmessagesdelayed{queue="queue-2",} 0.0
# HELP sqs_approximatenumberofmessages The approximate number of visible messages in a queue.
# TYPE sqs_approximatenumberofmessages gauge
sqs_approximatenumberofmessages{queue="queue-2",} 0.0

I created a deployment for this sqs-exporter using the container as instructed here. It seems that this exporter not get's discovered from my prometheus. Are there any clues to investigate why?

Thanks, in advance.

Hi, the exporter stopped working, can't see any sqs metrics getting scraped. Is this project still maintained?

Today we added our first FIFO sqs queue and we noticed our sqs-exporter stopped working. Apon checking the Prometheus targets (http://127.0.0.1:9090/targets) I could see the sqs-exporter servicemonitor was showing as down with the error message:

invalid metric type "fifo_sqs_approximatenumberofmessagesdelayed gauge"

As a workaround, we added the SQS_QUEUE_NAMES ENV var to our sqs-exporter deployment Manifest to whitelist the queues we want metrics for:

spec:
  containers:
  - name: sqs-exporter
  ...
  env:
  - name: SQS_QUEUE_NAMES
    value: "redacted_queue_name1,redacted_queue_name2,redacted_queue_name3"

This way, our new FIFO queue is ignored.

It is my understanding that FIFO queues seem to have to have a name the ends in .fifo (source). Maybe the . character in the queue name is causing a issue?

Please close this issue if I'm misunderstanding something - which is very possible.

Is this project still maintained? Also, is there any way to run the exporter without docker?

Hi,

I'm trying this export I have a problem on the queue detection.
I want to set manually which queues are monitored.
So I see the exporter support this environment variable: SQS_QUEUE_NAMES set to myQueue0,MyQueue1

But the exporter will fetch all queues ...
When I remove all list queues capabilities the exporter doesn't return metrics

I don't understand why I have this isssue because when I'm reading the code:

String queueNames = System.getenv("SQS_QUEUE_NAMES");
String queueNamePrefix = System.getenv("SQS_QUEUE_NAME_PREFIX");
if (queueNames != null) {
   // find the URLs for the named queues .
   String[] names = queueNames.split(",");
   queueUrls = new ArrayList<String>();
   for(String name : names) {
       queueUrls.add(sqs.getQueueUrl(name).getQueueUrl());
   }
} else {
    // get URLs for all queues visible to this account (with prefix if specified)
    ListQueuesResult queues = sqs.listQueues(queueNamePrefix); //If null is passed in the whole                     unfiltered list is returned
    queueUrls = queues.getQueueUrls();
}

It seems the code enter in the else instead of the if or I miss something (I'm not java dev)

version: jmal98/sqs-exporter:0.0.5
iam policy:

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "Stmt1542622727660",
      "Action": [
        "sqs:GetQueueAttributes",
        "sqs:GetQueueUrl",
        "sqs:ListDeadLetterSourceQueues",
        "sqs:ListQueueTags",
        "sqs:ListQueues"
      ],
      "Effect": "Allow",
      "Resource": "*"
    }
  ]
}

container logs:

19 Nov 2018 16:08:30,109 org.jmal98.metrics.collector.Sqs:44 AmazonSQS client is connected to region: (eu-west-1)
2018-11-19 16:08:31.964:INFO::main: Logging initialized @3733ms to org.eclipse.jetty.util.log.StdErrLog
2018-11-19 16:08:32.051:INFO:oejs.Server:main: jetty-9.4.z-SNAPSHOT; built: 2018-06-05T18:24:03.829Z; git: d5fc0523cfa96bfebfbda19606cad384d772f04c; jvm 1.8.0_171-b11
2018-11-19 16:08:32.112:INFO:oejs.AbstractConnector:main: Started ServerConnector@3574e198{HTTP/1.1,[http/1.1]}{0.0.0.0:9384}
2018-11-19 16:08:32.114:INFO:oejs.Server:main: Started @3882ms
19 Nov 2018 16:08:32,115 org.jmal98.sqs.exporter.Application:51 Exporter has started.

Are there plans to add fix for log4j vulnerability for this exporter

https://www.zdnet.com/article/apache-releases-new-2-17-0-patch-for-log4j-to-solve-denial-of-service-vulnerability/

FIrst, I really appreciate you putting this repo together! I had some challenges getting the metrics from this container into my k8s based prometheus. I could curl the service and it was working fine but not showing up in prometheus.

After much searching found out you have to annotate your deployments for prometheus to discover them. I'm pasting my simple example below in hopes you can include this type of info in your README.md Perhaps this is common knowledge but my sense is this would help noobs like me get working faster.

Deployment.yaml

kind: Deployment
metadata:
  name: sqs-exporter
  labels:
    app: sqs-exporter
spec:
  replicas: 1
  selector:
    matchLabels:
      app: sqs-exporter
  template:
    metadata:
      labels:
        app: sqs-exporter
      annotations:
        prometheus.io/scrape: "true"
        prometheus.io/path: /metrics
        prometheus.io/port: "9384"
    spec:
      containers:
      - name: sqs-exporter
        image: jmal98/sqs-exporter
        ports:
        - containerPort: 9384
        env:
          - name: AWS_ACCESS_KEY_ID
            valueFrom:
              secretKeyRef:
                name: batch-iam
                key: access-key
          - name: AWS_SECRET_ACCESS_KEY
            valueFrom:
              secretKeyRef:
                name: batch-iam
                key: secret

service.yaml

apiVersion: v1
kind: Service
metadata:
  name: sqs-exporter
  labels:
    app: sqs-exporter
spec:
  selector:
    app: sqs-exporter
  ports:
  - protocol: TCP
    port: 9384
    targetPort: 9384

I've noticed this after creating a new queue and not updating the exporter IAM role.
We had 4 queues:

1. qa-queue-cpu
2. qa-queue-gpu
3. qa-queue-mem
4. qa-queue-tasks

The exporter was configured with SQS_QUEUE_NAME_PREFIX=qa and was working fine, yesterday we added a new queue qa-queue-split and didn't update the IAM role for the exporter, then I noticed that the qa-queue-tasks metrics are gone, but all the other are working fine.
After updating the IAM role I finally got metrics from all the queues.

It seems to me that queues are being sorted alphabetically, and if the exporter doesn't have access to one of them, it'll stop at that one, this explains why it didn't get metrics from a queue that was working before we added the new one.