[MAGNOLIA-9250] Investigate publication slowness on PaaS test instance Created: 12/Jan/24  Updated: 08/Feb/24

Status: Selected
Project: Magnolia
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Spike Priority: Major
Reporter: Michael Duerig Assignee: Alberto Soto
Resolution: Unresolved Votes: 0
Labels: dx-core-6.3
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Attachments: PNG File image-2024-01-16-21-26-05-209.png     PNG File image-2024-01-16-21-29-02-134.png    
Issue Links:
dependency
Template:
Documentation update required:
Yes
Date of First Response:
Epic Link: Throughput improvements
Work Started:

 Description   
  • When publishing a node with 100 children, the server went down. Also when generating it. This does not happen when the server triggers 101 publications independently
  • Open 100 subnodes in the UI already takes long time
  • Document findings in our large scale content (MAGNOLIA-9083) guide if applicable


 Comments   
Comment by Michael Duerig [ 16/Jan/24 ]

Can we get some thread dumps and a heap histogram from the time when this happens?

Comment by Michael Duerig [ 22/Jan/24 ]

I assume MAGNOLIA-9176 should improve things here. The increase in memory usage is likely from the spike in number of threads, which that ticket should significantly lower.

Scratch that, on 6.3 we already run observation off a thread pool regardless of MAGNOLIA-9176. Guess we have to look at a thread dump to understand more.

Comment by Michael Duerig [ 22/Jan/24 ]

I quickly tried to reproduce this locally (root page with 100 pages each again with 100 pages). Nothing bad happened, my instance stayed stable and reactive, CPU, memory and thread count were all bounded and reasonable.

Generated at Mon Feb 12 04:39:58 CET 2024 using Jira 9.4.2#940002-sha1:46d1a51de284217efdcb32434eab47a99af2938b.