[
https://issues.apache.org/jira/browse/CLOUDSTACK-9405?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15316553#comment-15316553
]
dsclose commented on CLOUDSTACK-9405:
-------------------------------------
Steps to (easily) reproduce:
1. Authenticate on a Cloudstack management service.
2. Browse to the URL at /client/api?response=json&command=listDomains
This takes ~26.9s on Cloudstack 4.8.0. On Cloudstack 4.5.2 it takes < 3s.
Examining the database during the API request, I see that there an increase in
load on the database during the API call. This increase on load is, however,
extremely small.
> listDomains API call takes an extremely long time to respond
> ------------------------------------------------------------
>
> Key: CLOUDSTACK-9405
> URL: https://issues.apache.org/jira/browse/CLOUDSTACK-9405
> Project: CloudStack
> Issue Type: Bug
> Security Level: Public(Anyone can view this level - this is the
> default.)
> Components: API
> Affects Versions: 4.8.0
> Reporter: dsclose
> Labels: performance
>
> We recently upgraded from Cloudstack 4.5.2 to Cloudstack 4.8.0. Since this
> update, the listDomains API call has started taking an extremely long time to
> respond. This has caused issues with our services that rely on this API call.
> Initially they simply timed out until we increased the thresholds. Now we
> have processes that used to take a few seconds taking many minutes.
> This is so problematic for us that our organisation has put a halt on further
> updates of Cloudstack 4.5.2 installations. If reversing the update of zones
> already on 4.8.0 was feasible, we would have reverted back to 4.5.2.
> Here is a table of the times we're seeing:
> ||CS Version||Domain Count||API Response Time||
> |4.5.2|251|~3s|
> |4.8.0|182|~26s|
> |4.8.0|<10|<1s|
> This small data sample indicates that the response time for zones with a
> larger amount of domains is significantly worse after the update to 4.8.0.
> Zones with few domains aren't able to reproduce this issue.
> I recall a bug being resolved recently that concerned reducing the response
> time for list* API calls. I also recall [~remibergsma] resolving a bug
> concerning the sorting of the listDomains response. Is it possible that these
> issues are connected?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
