[ https://issues.apache.org/jira/browse/CLOUDSTACK-9377?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15281247#comment-15281247 ]
Rohit Yadav commented on CLOUDSTACK-9377: ----------------------------------------- I had a quick look at the code, the issue indeed will happen if a cluster has more than 500 (default page size) the issue you've mentioned will occur. A quick hack would be to pass pagesize=-1 in the list APIs, for example can you test by placing pagesize=-1 here https://github.com/apache/cloudstack/blob/master/ui/scripts/metrics.js#L166 and comment if that fixes the issue for you. Practically, I would assume nobody would have more than 500 hosts in a cluster but I don't know if we've users out there having something like that. > Metrics data incorrectly calculated in zone/cluster metrics view > ---------------------------------------------------------------- > > Key: CLOUDSTACK-9377 > URL: https://issues.apache.org/jira/browse/CLOUDSTACK-9377 > Project: CloudStack > Issue Type: Bug > Security Level: Public(Anyone can view this level - this is the > default.) > Components: UI > Affects Versions: 4.7.0 > Reporter: Rashmi Dixit > Labels: UI, metrics > > In the zone metrics view, listClusters and listHosts APIs are being used to > compute average used and allocated data for metrics calculations. > Issue: > 1. All APIs have a default pagesize. Count of results are always equal to the > default pagesize if not explicitly passed. The UI is not passing anything > related to pagesize to these APIs > 2. Example to explain issue > 600 hosts in a cluster. Default Page size = 500. > There is only one call to fetch clusters. This will fetch a maximum of 500 > records. If there are more, then these aren't fetched. > For each cluster, listHosts is called to fetch hosts. Since page size is 500 > only 500 results are returned. > Computation for 500 hosts is done but for calculating average this value is > divided by 600. This is incorrect. > UI should fetch all hosts in one go and correctly calculate metrics. However, > this will result in performance degradation as number of hosts/clusters > increase. > Ideally a new API needs to be introduced for this metrics calculation. > Same issue is present in the cluster metrics view. -- This message was sent by Atlassian JIRA (v6.3.4#6332)