-
Bug
-
Resolution: Obsolete
-
Medium
-
None
-
2.10
Indexing takes 60-90 minutes on EAC, which is a lot slower than it was before. We need to figure out why that is.
[CONFSERVER-12943] Indexing apparently slower on EAC
Per Fragemann [Atlassian]
added a comment - I can't see us doing any work on this right now, removing fix-version 3.0. At least we have the performance-test to prevent any other regression in the future.
Per Fragemann [Atlassian]
added a comment - I can't see us doing any work on this right now, removing fix-version 3.0. At least we have the performance-test to prevent any other regression in the future. Per Fragemann [Atlassian]
added a comment - We should at least have George create an indexing- loadtest for us. If we still have a little time left during the 3.0 release cycle we may still have time for a quick look (I doubt it though). Assigning to Matt so the loadtest is on his radar as well.
Per Fragemann [Atlassian]
added a comment - We should at least have George create an indexing- loadtest for us. If we still have a little time left during the 3.0 release cycle we may still have time for a quick look (I doubt it though). Assigning to Matt so the loadtest is on his radar as well. Some pretty rough observations from the recent upgrade to 2.10 on EAC.
The re-index took about 90 minutes to complete.
Both nodes would regularly bounce between 80% - 370% CPU utilitization.
The DB server, even with both nodes indexing at the same time was rarely jumping above 90% CPU Utilization.
Now if you want a stranger story here's one. node1 finished indexing first. However it's reindexing progress screen started showing the progress of node 2. Checking the screen on either node would get you the same results until node 2 finished.
Dave L can confirm. I'm not making this sh*t up.
Per Fragemann [Atlassian]
added a comment - We have not been able to look at this during the 2.10 release cycle, but we must check it for 3.0 in the ER team.
Per Fragemann [Atlassian]
added a comment - We have not been able to look at this during the 2.10 release cycle, but we must check it for 3.0 in the ER team. Per Fragemann [Atlassian]
added a comment - No progress on this issue for a while. Can we make sure this gets addressed before we spend more time on new features and improvements?
Per Fragemann [Atlassian]
added a comment - No progress on this issue for a while. Can we make sure this gets addressed before we spend more time on new features and improvements? 
This issue is now obsolete. Changes are underway in the re-indexing code to address many of the issues in the re-indexing process. CONFDEV-5127 tracks that work.