SOLVED

Availability on OMS

Brass Contributor

Hi everyone.

I'm trying to find a way of getting Availability of servers on OMS, but I can't find any...

By Availability I mean the % of uptime of a given server during a certain period of time.

So, if a server was up 98 of a total 100 hours, the availability for that period is 98%.

I'm looking to do that in OMS, but I'm not sure it's possible.

Thanks in advance.

36 Replies
Hi There is no out of the box solution for this. Probably one way to calculate this is by using Heartbeat events. For Windows machines they are logged every 1 minute and for Linux every 5 minutes. You could probably calculate what should be the sum of the events for each machine and also the maximum sum for specific time frame. From those two values you can calculate the percentage. The query language is pretty rich so this should be possible.

Stanislav is right, it's possible :)

Here's an example that calculates the availability rate of each computer, starting at midnight.

let midnight=startofday(now());
Heartbeat
| where TimeGenerated>midnight
| summarize heartbeat_per_hour=count() by bin_at(TimeGenerated, 1h, midnight), Computer
| extend available_per_hour=iff(heartbeat_per_hour>0, true, false)
| summarize total_available_hours=countif(available_per_hour==true) by Computer 
| extend number_of_buckets=hourofday(now())+1
| extend availability_rate=total_available_hours*100/number_of_buckets

Run it on our playground and tweak it as makes sense to you.

Thank you for your help, im going to investigate a bit that query.

However, i'm not sure about that approach because the heartbeat happens to stop working a lot even if the VM is perfectly fine.

But I understand the approach...

 

Heartbeat should be running without issue however there are scenarios when you might not get data: - Log Analytics service is down - You machine has lost Internet connection - MMA agent service is stopped - MMA agent is not functioning properly

The last situation MMA agent not working properly or has stopped working is exactly what worries me in order to create the availability report based on heartbeats.

Noa, your script is amazing, however i'm struggling to understand it and tweak it to my needs (30 fixed days, for example from 1st to 31 of january)

Could you gimme a hand to understand it?

 

let midnight=startofday(now()) ; #First part. I need to change this to between((2018-01-01) .. (2017-01-31)); am I correct?
Heartbeat
| where TimeGenerated>midnight
| summarize heartbeat_per_hour=count() by bin_at(TimeGenerated, 1h, midnight) #im not sure i understand why do you use bin_at instead of just bin, Computer
| extend available_per_hour=iff(heartbeat_per_hour>0, true, false)
| summarize total_available_hours=countif(available_per_hour==true) by Computer 
| extend number_of_buckets=hourofday(now())+1
| extend availability_rate=total_available_hours*100/number_of_buckets

 

best response confirmed by Dante Nahuel Ciai (Brass Contributor)
Solution

Sure. I tweaked it a bit to match what you ask for: 

let start_time=startofday(datetime("2017-01-01"));
let end_time=endofday(datetime("2017-01-31"));
Heartbeat
| where TimeGenerated > start_time and TimeGenerated < end_time
| summarize heartbeat_per_hour=count() by bin_at(TimeGenerated, 1h, start_time), Computer
| extend available_per_hour=iff(heartbeat_per_hour>0, true, false)
| summarize total_available_hours=countif(available_per_hour==true) by Computer 
| extend total_number_of_buckets=round((end_time-start_time)/1h)
| extend availability_rate=total_available_hours*100/total_number_of_buckets

The first 2 lines define variables, set to the start and end time you mentioned.

 

Next, we use these variables to limit the query to that time range: 

| where TimeGenerated > start_time and TimeGenerated < end_time

Then we count the heartbeats reported from each computer, in buckets (bins) of 1 hour, starting at the start time you define: 

| summarize heartbeat_per_hour=count() by bin_at(TimeGenerated, 1h, start_time), Computer

Now we can see how many heartbeats were reported by each computer each hour. If the number is  0 we understand the computer was probably offline at that time.

We use a new column to mark if a computer was available or not each hour: 

| extend available_per_hour=iff(heartbeat_per_hour>0, true, false)

and then count the number of hours each computer was indeed "alive": 

| summarize total_available_hours=countif(available_per_hour==true) by Computer

Note that this way we give a little leeway for missing heartbeat reports each hour. Instead of expecting a report every 5 or 10 minutes, we only mark a computer as "unavailable" if we didn't get any report from it during a full hour.

 

At this point we get a number for each computer, something like this: 

results1.png

 

So we know each computer was alive 11 hours in the select time range. But what does it mean? how many hours were there altogether? is this 11 out of 11 hours (100% availability) or out of 110 hours (only 10% availability)?

Here's how we can calculate the total number of hours in the selected time range: 

| extend total_number_of_buckets=round((end_time-start_time)/1h)+1

I admit it might not be the best calculation of buckets.. there is probably a better way but I can't think of it now..

 finally we calculate the ratio between available hours and total hours:

| extend availability_rate=total_available_hours*100/total_number_of_buckets

and get this:

results2.png

 

HTH,

Noa

Noa, amazing, thank you so much.

Excellent query. How can "let midnight=startofday(now())" be altered to make it my local time zone? If I run this as is, it seems to be my time +7, and the amount of hours don't match up.

Hi,

 

Thanks for a exellent code sample.

 

I would like to extend the Query, supporting also specified time intervals and smaller uptime checks (heartbeat)

# Service levels

Ex: Service agreements are based on 3 categories

S1 = 07:00 - 17:00 Weekdays

S2 = 07:00- 22:00 Weekdays

365/7 = Always (already supported by your query

 

= Uptime should be calculated based on service agreement hours/days

 

Time should also be converted to UTC +1

- will this do the trick = >  

Heartbeat
| extend Timegenerated = TimeGenerated + 1h
 
I checked the samples from endofday/week, but are unable to get it to work in your sample
 
 
# Intervals
extend available_per_hour=iff(heartbeat_per_hour>0, true, false)
= can this be adjusted to heartbeat per 30 min / 15 min
 
 
Any ideas ?
 
Br
erik

Thanks George.

To adjust for the local time zone you can do this:

let midnight=startofday(now())-7h

Hi Eric,

 

To adjust for the service agreement, you can calculate the start time and end time like this:

let raw_date = datetime("2017-01-01");
let start_date = case("SLA" in ("S1", "S2"), case(dayofweek(raw_date)==0, startofday(raw_date+1d)+7h,
                                            dayofweek(raw_date)==6, startofday(raw_date+2d)+7h, 
                                            startofday(raw_date)+7h),
                raw_date);

On the intervals - it can adjusted any way you need, just use `bin(fieldname, 30m)` instead of `bin(fieldname, 1h)`.

it is a pwerShell Script?

Hey,

That's not powershell but our query language, that can be used here and through our API.

I am struggling to generate the report for Mon-Friday only and in my time zone. I just get errors.

 

let start_time=startofday(datetime("2018-07-1 07:30:00"));
let end_time=endofday(datetime("2018-07-31 18:00:00"));
Heartbeat
| where TimeGenerated > start_time and TimeGenerated < end_time
| summarize heartbeat_per_hour=count() by bin_at(TimeGenerated, 1h, start_time), Computer
| extend available_per_hour=iff(heartbeat_per_hour>0, true, false)
| summarize total_available_hours=countif(available_per_hour==true) by Computer
| extend total_number_of_buckets=round((end_time-start_time)/1h)
| extend availability_rate=total_available_hours*100/total_number_of_buckets

I am struggling to generate the report for Mon-Friday only and in my time zone. I just get errors. The script below works for me. Thanks

 

let start_time=startofday(datetime("2018-07-1 07:30:00"));
let end_time=endofday(datetime("2018-07-31 18:00:00"));
Heartbeat
| where TimeGenerated > start_time and TimeGenerated < end_time
| summarize heartbeat_per_hour=count() by bin_at(TimeGenerated, 1h, start_time), Computer
| extend available_per_hour=iff(heartbeat_per_hour>0, true, false)
| summarize total_available_hours=countif(available_per_hour==true) by Computer
| extend total_number_of_buckets=round((end_time-start_time)/1h)
| extend availability_rate=total_available_hours*100/total_number_of_buckets
Hey Noa,

Can we take the 1 year details by this script.?

Can  we Availabilty for past 10 days instead of add start date and End date

Thanks but I got the answer

let month = startofday(ago(3d));
Heartbeat
| where TimeGenerated>ago(3d)
| summarize heartbeat_per_hour=count() by bin_at(TimeGenerated, 1h, (ago(3d))), Computer
| extend available_per_hour=iff(heartbeat_per_hour>0, true, false)
| summarize total_available_hours=countif(available_per_hour==true) by Computer
| extend total_number_of_buckets= round((now()-month)/1h)-2
| extend availability_rate=total_available_hours*100/total_number_of_buckets
1 best response

Accepted Solutions
best response confirmed by Dante Nahuel Ciai (Brass Contributor)
Solution

Sure. I tweaked it a bit to match what you ask for: 

let start_time=startofday(datetime("2017-01-01"));
let end_time=endofday(datetime("2017-01-31"));
Heartbeat
| where TimeGenerated > start_time and TimeGenerated < end_time
| summarize heartbeat_per_hour=count() by bin_at(TimeGenerated, 1h, start_time), Computer
| extend available_per_hour=iff(heartbeat_per_hour>0, true, false)
| summarize total_available_hours=countif(available_per_hour==true) by Computer 
| extend total_number_of_buckets=round((end_time-start_time)/1h)
| extend availability_rate=total_available_hours*100/total_number_of_buckets

The first 2 lines define variables, set to the start and end time you mentioned.

 

Next, we use these variables to limit the query to that time range: 

| where TimeGenerated > start_time and TimeGenerated < end_time

Then we count the heartbeats reported from each computer, in buckets (bins) of 1 hour, starting at the start time you define: 

| summarize heartbeat_per_hour=count() by bin_at(TimeGenerated, 1h, start_time), Computer

Now we can see how many heartbeats were reported by each computer each hour. If the number is  0 we understand the computer was probably offline at that time.

We use a new column to mark if a computer was available or not each hour: 

| extend available_per_hour=iff(heartbeat_per_hour>0, true, false)

and then count the number of hours each computer was indeed "alive": 

| summarize total_available_hours=countif(available_per_hour==true) by Computer

Note that this way we give a little leeway for missing heartbeat reports each hour. Instead of expecting a report every 5 or 10 minutes, we only mark a computer as "unavailable" if we didn't get any report from it during a full hour.

 

At this point we get a number for each computer, something like this: 

results1.png

 

So we know each computer was alive 11 hours in the select time range. But what does it mean? how many hours were there altogether? is this 11 out of 11 hours (100% availability) or out of 110 hours (only 10% availability)?

Here's how we can calculate the total number of hours in the selected time range: 

| extend total_number_of_buckets=round((end_time-start_time)/1h)+1

I admit it might not be the best calculation of buckets.. there is probably a better way but I can't think of it now..

 finally we calculate the ratio between available hours and total hours:

| extend availability_rate=total_available_hours*100/total_number_of_buckets

and get this:

results2.png

 

HTH,

Noa

View solution in original post