Styjun

Are randomly-generated passwords starting with "a" less secure?

Is there a way to perform "if" check only once instead of every iteration in a loop?

Where or how can I find what interfaces an out of the box Apex class implements?

Was the Ford Model T black because of the speed black paint dries?

Credit union holding car note, refuses to provide details of how payments have been applied

Steampunk helicopter

Is purchasing foreign currency before going abroad a losing proposition?

QGIS Welcome page: What is 'pin to list' for?

Do native speakers use ZVE or CPU?

Crowbar circuit causes unexpected behavior for op amp circuit

affect vs effect

Was adding milk to tea started to reduce employee tea break time?

(async () => )(); what is this?

What would the EU do if an EU member declared war on another EU member?

Why are Hobbits so fond of mushrooms?

What would be the ideal melee weapon made of "Phase Metal"?

Keep milk (or milk alternative) for a day without a fridge

Dropping outliers based on "2.5 times the RMSE"

Can I call 112 to check a police officer's identity in the Czech Republic?

Referring to different instances of the same character in time travel

What is temperature on a quantum level

What is the difference between logical consistency and logical entailment in deductive logic?

Is an acid a salt or not?

Why did my rum cake turn black?

Struggling with a Prometheus percentile query on jvm stats

Prometheus - Docker/JVM MonitoringPrometheus Startup queryPrometheus nested alert query in prometheusQuery span duration in PrometheusPrometheus query returns invalid valueHow to measure HTTP Latency with Percentile in PrometheusAverage Memory Usage Query - PrometheusPrometheus - debugging slow queryHow to monitor pool stats in Atomikos with prometheusPrometheus/Grafana plot wait time percentiles for jobs

.everyoneloves__top-leaderboard:empty,.everyoneloves__mid-leaderboard:empty,.everyoneloves__bot-mid-leaderboard:empty margin-bottom:0;

0

I'm writing a query to get the 95th percentile on the rate of change in the jvm_gc_collection_seconds_count. I want to trigger an alert when there is an unexpected spike. The purpose is to send me an alert when the JVM is working harder than usual.

I am using jvm_memory_bytes_used for testing as the rate of the GC does not show any anomalies right now. Below is an example of the data I get back from the jvm_memory_bytes_used. Should the first top two blue spikes not be returned as data points? The bottom red line should surely be picked up as an anomaly.

I have calculated the sum of the rate over 30 minutes grouped by instance. With that I try to get the 95th percentile.

histogram_quantile(0.95, sum(rate(jvm_memory_bytes_used[30m])) by (instance, le))

I've tried lowering the percentile which has no effect.

1. Am I using the correct variable to calculate the GC stats?


2. How can I get the histogram_quantile to produce results?


3. Am I taking the right approach using a percentile to achieve my goal?

prometheus

share|improve this question

asked Mar 26 at 4:26

cbillowescbillowes

2,01822 gold badges2121 silver badges3939 bronze badges

add a comment | 

0

I'm writing a query to get the 95th percentile on the rate of change in the jvm_gc_collection_seconds_count. I want to trigger an alert when there is an unexpected spike. The purpose is to send me an alert when the JVM is working harder than usual.

I am using jvm_memory_bytes_used for testing as the rate of the GC does not show any anomalies right now. Below is an example of the data I get back from the jvm_memory_bytes_used. Should the first top two blue spikes not be returned as data points? The bottom red line should surely be picked up as an anomaly.

I have calculated the sum of the rate over 30 minutes grouped by instance. With that I try to get the 95th percentile.

histogram_quantile(0.95, sum(rate(jvm_memory_bytes_used[30m])) by (instance, le))

I've tried lowering the percentile which has no effect.

1. Am I using the correct variable to calculate the GC stats?


2. How can I get the histogram_quantile to produce results?


3. Am I taking the right approach using a percentile to achieve my goal?

prometheus

share|improve this question

asked Mar 26 at 4:26

cbillowescbillowes

2,01822 gold badges2121 silver badges3939 bronze badges

add a comment | 

I'm writing a query to get the 95th percentile on the rate of change in the jvm_gc_collection_seconds_count. I want to trigger an alert when there is an unexpected spike. The purpose is to send me an alert when the JVM is working harder than usual.

I am using jvm_memory_bytes_used for testing as the rate of the GC does not show any anomalies right now. Below is an example of the data I get back from the jvm_memory_bytes_used. Should the first top two blue spikes not be returned as data points? The bottom red line should surely be picked up as an anomaly.

I have calculated the sum of the rate over 30 minutes grouped by instance. With that I try to get the 95th percentile.

histogram_quantile(0.95, sum(rate(jvm_memory_bytes_used[30m])) by (instance, le))

I've tried lowering the percentile which has no effect.

1. Am I using the correct variable to calculate the GC stats?


2. How can I get the histogram_quantile to produce results?


3. Am I taking the right approach using a percentile to achieve my goal?

prometheus

share|improve this question

asked Mar 26 at 4:26

cbillowescbillowes

2,01822 gold badges2121 silver badges3939 bronze badges

histogram_quantile(0.95, sum(rate(jvm_memory_bytes_used[30m])) by (instance, le))

prometheus

prometheus

share|improve this question

asked Mar 26 at 4:26

cbillowescbillowes

2,01822 gold badges2121 silver badges3939 bronze badges

share|improve this question

asked Mar 26 at 4:26

cbillowescbillowes

2,01822 gold badges2121 silver badges3939 bronze badges

asked Mar 26 at 4:26

cbillowescbillowes

2,01822 gold badges2121 silver badges3939 bronze badges

asked Mar 26 at 4:26

cbillowescbillowes

2,01822 gold badges2121 silver badges3939 bronze badges