Cluster high load unexpected

The Enterprise Cloud DB Monitoring Console shows high cluster load (see attachment). The actual load on the system that connects to the system was none. I have checked all of the openned DB connections and all were idle. No backups were active.
The load can go up and stay high, can someone suggest how to investigate futher ?

Best Regards,
Aleksandar

Its possible this forum post may provide some insight: http://forums.enterprisedb.com/posts/list/3205.page#12014

Thank you for the reply! I guess it should not be related to the backups since they run once a day, and the load peaks are more than one (as visible on the attached screenshot). Also I'm not sure how long a backup takes. Do you think it could not complete in hour (the first peak) - it is about very few data, the problematic cluster has 2G reserved with only 18% occupied data space?

Best Regards,
Aleksandar

Please check your Events starting before the first spike and continuing past the second spike looking for an event close to the time of each spike that may give you a clue. You can also copy the events and post them here if you like.

This message was edited 1 time. Last update was at 11 May 2012 10:46:25

Already have done it, I can't see something suspicious for the day :

All are within 10minutes. Same events for the previous and next day.

Best Regards,
Aleksandar

1. Please post the approximate times of the spikes

2. Please post times for the events (you only have the event text)

3. Please post all the events/times starting before the first spike and continuing past the second spike

Well, from the stat with resolution "last 24h" it's hard to tell the exact start and end times...
The second peak matches the backup window and continues about hour until fully calm. On the other side the first one last for hours with varying load ...
Is there a way to get more detailed statistic about the load in the past ?

Aleksandar

Forgot to ask, are you also running the m1.small size instance?

m1.small, yes

Best Regards,
Aleksandar

m1.small, yes

Okay, I think that may explain it. We have found in our use that the m1.small is pretty 'under-powered' to handle any significant load and does experience spikes during backup. I also think you had up to 20 connections running on that instance which will also push the limits of that instance size. That instance size appears to be best suited for simple development tasks and functionality testing and not appropriate for high loads.