[ZBX-26182] Zabbix Server Performance Degradation After Update to Version 6.0.10 Created: 2025 Mar 15  Updated: 2025 Mar 16  Resolved: 2025 Mar 16

Status: Closed
Project: ZABBIX BUGS AND ISSUES
Component/s: Server (S)
Affects Version/s: 7.0.10
Fix Version/s: None

Type: Problem report Priority: Trivial
Reporter: Vinicius Freitas Assignee: Zabbix Support Team
Resolution: Won't fix Votes: 0
Labels: performance
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified
Environment:

OS: Redhat 8 (vcpu 6 / mem 25gb)
DB: Mariadb 6.10.21 (OS: el8 / vcpu 8 / mem 96gb / ssd 8T)
NVPS: ~9k
Proxies Count: 80

Number of hosts (enabled/disabled) 10864 10337 / 527
Number of templates 1352
Number of items (enabled/disabled/not supported) 2715946 1943367 / 601862 / 170717
Number of triggers (enabled/disabled [problem/ok]) 854963 791361 / 63602 [6645 / 784716]
Number of users (online) 104 5
Required server performance, new values per second 9026.65
High availability cluster Enabled Fail-over delay: 1 minute


Attachments: Text File server_diaginfo_6.0.10.txt     Text File server_diaginfo_after_rollback_15-mar.txt     PNG File zbx_cache_before-and-at_update.png     PNG File zbx_cache_graph_afeter_rollback.png     PNG File zbx_queue_before-and-at_update.png     PNG File zbx_queue_graph_afeter_rollback.png    

 Description   

Steps to reproduce:

  1. Update from 7.0.9 to 7.0.10

Result:

After the version update, our server, which has around 9k NVPS, experienced a significant delay in normalizing data synchronization with the proxies and sending it to the database. Even after 48 hours, the issues persisted. It was observed that the history syncer processes were barely "working"—it was very difficult to see them actually sending data to the database.

Another point noted was that the History Cache usage never dropped below 80% throughout this entire period. It's important to highlight that before the update to version 6.0.9, these issues were not present.

No changes were made apart from the minor version update to 6.0.10 on 03/12/2025.

Additionally, another factor observed was that the proxy queue, which was typically stabilized between 10k and 15k, spiked to over 1.5 million. However, after rolling back to version 6.0.9, everything returned to normal in less than an hour.

Support ticket: GOB-432

Expected:
The expectation was that normalization would occur within a maximum of 3 hours after the update and that the queue and history cache levels would return to normal.



 Comments   
Comment by Alexander Vladishev [ 2025 Mar 15 ]

Good afternoon! Please clarify the server version. The problem description is contradictory. In one place, it mentions 6.0.9 and 6.0.10, while in another, it states 7.0.9 and 7.0.10.

Comment by Vinicius Freitas [ 2025 Mar 16 ]

Sorry, I wrote it wrong, it's version 7, but I can't close this one I opened, so I opened another one correctly: ZBX-26184.

Comment by dimir [ 2025 Mar 16 ]

Closing on behalf of previous comment.

Generated at Fri Apr 04 17:42:32 EEST 2025 using Jira 9.12.4#9120004-sha1:625303b708afdb767e17cb2838290c41888e9ff0.