Stable means for counting running instances

[lee212]

The real time reports via the metrics framework is using euca-describe-instances to do reporting running instances but sometimes, it is not working properly due to its own errors or unavailability. To fix this type of possible issues, the framework needs to use its own database to count running instances with the euca tools. I believe this way would fix the issues that you were told about.

For example, time dismatch with local machine and the server generates errors like below: $ euca-describe-instances verbose 403 Forbidden: Message was generated in the future (times in UTC): Timestamp=2012-12-03T23:53:47Z

Executing 'ntpupdate ntp time server' will fix the problem but ntpupdate can be run by only root.

Here is the diagram for the proposed solution to mitigate this type of problems.

1. run euca-describe-instances
2. check the return value
3. if it is okay then use the return value
4. if it gets errors, use the framework to count real-time metrics

[laszewsk]

we need to augment the view with some kind of indication when such errors occur

Also I have seen first hand that Koji was able to see instances running while the metrics framework did not show them.

I am not sure if he used describe images or something else.

On Dec 3, 2012, at 7:50 PM, lee212 notifications@github.com wrote:

The real time reports via the metrics framework is using euca-describe-instances to do reporting running instances but sometimes, it is not working properly due to its own errors or unavailability. To fix this type of possible issues, the framework needs to use its own database to count running instances with the euca tools. I believe this way would fix the issues that you were told about.

For example, time dismatch with local machine and the server generates errors like below: $ euca-describe-instances verbose 403 Forbidden: Message was generated in the future (times in UTC): Timestamp=2012-12-03T23:53:47Z

Executing 'ntpupdate ntp time server' will fix the problem but ntpupdate can be run by only root.

Here is the diagram for the proposed solution to mitigate this type of problems.

run euca-describe-instances check the return value if it is okay then use the return value if it gets errors, use the framework to count real-time metrics — Reply to this email directly or view it on GitHub.