OpenNebula

This bug reappears in opennebula 4.0.0 and it looks like the original code quoted could be a problem.

When an user dosn't have any VM running its VM quotas are set to 0, what means unlimited instead of -1.
We should check this logic better to make it work with unlimited and default quotas.

fsck should solve problems, not create new one.
We had to solve it in 4.0.1.

It is not solved.
If I understand correctly all your changes, now if an user doesn't use any resources, then his custom quotas are replaced by default quotas, what is not right.
Probably it would be better to replace quota values by current quota values if they exists, then by default quotas.

Rolandas Naujikas wrote:

It is not solved.
If I understand correctly all your changes, now if an user doesn't use any resources, then his custom quotas are replaced by default quotas, what is not right.
Probably it would be better to replace quota values by current quota values if they exists, then by default quotas.

Can you point to the lines where this is happening?

Looking for example at the Image quotas section, if the user doesn't have any Image in use:

The first loop will go through the existing quotas. If a quota has a RVMS_USED different to 0, then it will be replaced with 0. The current quota limits (RVMS) are not replaced in any case.

        img_quota.each_element("IMAGE") { |img_elem|
            img_id = img_elem.get_text("ID").to_s

            rvms = img_usage.delete(img_id)

            rvms = 0 if rvms.nil?

            img_elem.each_element("RVMS_USED") { |e|
                if e.text != rvms.to_s
                    log_error("#{resource} #{oid} quotas: Image #{img_id}\tRVMS has #{e.text} \tis\t#{rvms}")
                    e.text = rvms.to_s
                end
            }
        }

The second loop will do nothing, since the user is not using any image. But if he were using images, this loop iterates over the images that are in use, but don't have a quota reflecting this usage currently. So the RVMS = "-1" is not replacing any existing quota limit here either.

        img_usage.each { |img_id, rvms|
            log_error("#{resource} #{oid} quotas: Image #{img_id}\tRVMS has 0 \tis\t#{rvms}")

            new_elem = img_quota.add_element("IMAGE")

            new_elem.add_element("ID").text = img_id
            new_elem.add_element("RVMS").text = "-1" 
            new_elem.add_element("RVMS_USED").text = rvms.to_s
        }

Regards

It's about original bug description lines, related to CPU, MEMORY, VMS quotas.
I didn't check about others, probably they are OK.

I just tested this version of fsck.rb against our testbed on 4.0.1 (with version changed to 4.0.1).
It looks much better and those accounts, with missing quotas (as in sunstone 4.0.1 when I try set to default) after "onedb fsck" it is changed to default quotas.
Sorry for my mistake. Before saying something, it is better to test it.

Rolandas Naujikas wrote:

I just tested this version of fsck.rb against our testbed on 4.0.1 (with version changed to 4.0.1).
It looks much better and those accounts, with missing quotas (as in sunstone 4.0.1 when I try set to default) after "onedb fsck" it is changed to default quotas.
Sorry for my mistake. Before saying something, it is better to test it.

Great! Thanks for keeping an eye on our commits.